Search

Generating fluent, coherent, and informative text from structured data is called table-to-text generation. Copying words from the table is a common method to solve the “out-of-vocabulary” problem, but it’s difficult to achieve accurate copying. In order to overcome this problem, we invent an auto-regressive framework based on the transformer that combines a copying mechanism and language modeling to generate target texts. Firstly, to make the model better learn the semantic relevance between table and text, we apply a word transformation method, which incorporates the field and position information into the target text to acquire the position of where to copy. Then we propose two auxiliary learning objectives, namely table-text constraint loss and copy loss. Table-text constraint loss is used to effectively model table inputs, whereas copy loss is exploited to precisely copy word fragments from a table. Furthermore, we improve the text search strategy to reduce the probability of generating incoherent and repetitive sentences. The model is verified by experiments on two datasets and better results are obtained than the baseline model. On WIKIBIO, the result is improved from 45.47 to 46.87 on BLEU and from 41.54 to 42.28 on ROUGE. On ROTOWIRE, the result is increased by 4.29% on CO metric, and 1.93 points higher on BLEU.

Multi-keyword query is widely supported in text search engines. However, an analogue in image retrieval systems, multi-object query, is rarely studied. Meanwhile, traditional object-based image retrieval methods often involve multiple steps separately. In this work, we propose a weakly-supervised Deep Multiple Instance Hashing (DMIH) approach for multi-object image retrieval. Our DMIH approach, which leverages a popular CNN model to build the end-to-end relation between a raw image and the binary hash codes of its multiple objects, can support multi-object queries effectively and integrate object detection with hashing learning seamlessly. We treat object detection as a binary multiple instance learning (MIL) problem and such instances are automatically extracted from multi-scale convolutional feature maps. We also design a conditional random field (CRF) module to capture both the semantic and spatial relations among different class labels. For hashing training, we sample image pairs to learn their semantic relationships in terms of hash codes of the most probable proposals for owned labels as guided by object predictors. The two objectives benefit each other in a multi-task learning scheme. Finally, a two-level inverted index method is proposed to further speed up the retrieval of multi-object queries. Our DMIH approach outperforms state-of-the-arts on public benchmarks for object-based image retrieval and achieves promising results for multi-object queries.

Natural language processing (NLP) is a field of computer science concerned with automated text and language analysis. In recent years, following a series of breakthroughs in deep and machine learning, NLP methods have shown overwhelming progress. Here, we review the success, promise and pitfalls of applying NLP algorithms to the study of proteins. Proteins, which can be represented as strings of amino-acid letters, are a natural fit to many NLP methods. We explore the conceptual similarities and differences between proteins and language, and review a range of protein-related tasks amenable to machine learning. We present methods for encoding the information of proteins as text and analyzing it with NLP methods, reviewing classic concepts such as bag-of-words, k-mers/n-grams and text search, as well as modern techniques such as word embedding, contextualized embedding, deep learning and neural language models. In particular, we focus on recent innovations such as masked language modeling, self-supervised learning and attention-based models. Finally, we discuss trends and challenges in the intersection of NLP and protein research.

The users helped by the search engine for online transactions in e-commerce; nonetheless, there is still a lack of search interest from the user and online shopping intentions. To boost the user's search for product recommendations, use a search engine for the quest but not for purchasing purposes. Existing frameworks have some significant problems in the recommendation technology, such as new product problems, fewer evaluation problems, a vast amount of data, etc. Many e-commerce applications lack a better search experience like an Exploratory search system. This research work aims to create a conceptual model for recommendation systems using exploratory search, to study the behavior of users and the efficacy of exploratory search in terms of the quality of the search results produced. Several machine learning algorithms are used in this research to classify e-commerce products and evaluate the performance of these algorithms. The full-text search mechanism is used to implement an exploratory search system. The exploratory search system is evaluated by three criteria called to look up, learn and investigate. After the experiments and evaluation, it is observed that AdaBoost over decision tree performs better than the other classification algorithms implemented. The exploratory search system satisfies the three requirements that are lookup learn and investigate during the search process. Contributions to this work are to build a conceptual model for the recommendation system through the dataset of user events and finding the implementation of an exploratory search mechanism in an e-commerce application.

Semantic network model for representing data and knowledge was analysed. Selection of this model for working with text information was justified. The objective of automatic semantic network generation based on an arbitrary Russian-language text was formulated. Initial data, conditions and constraints necessary for network generation algorithm are listed. As a result of the part-of-speech analysis for each word and word order in a sentence, semantic relations between words are determined. The Lexeme dictionary was created to determine the part of speech of words in sentences. A set of question types used in the semantic network was selected. The number of relations in the network is regulated due to the possibility to use only necessary relation types when resolving a specific task. With that, the relations in semantic network can have very different types, which makes it a universal model for representing data and knowledge. The algorithm was developed which allows one to get answers for the questions asked. The semantic network model was generated automatically for the sentences considered. In the proposed algorithm the semantic network is interpreted as unoriented graph on which breadth-first search algorithm is used to find an answer. The proposed algorithms were implemented in a software tool which automatically generates the semantic network for an arbitrary text. The created software tool allows asking questions and getting answers to them based on the information which is stored in the semantic network. The experiments have shown that the generated semantic network gives correct answers to the questions posed. The network is modified by adding and removing information in it. There is a possibility to choose complexity of network structure depending on a specific task being resolved. The proposed approach for building and working with the semantic network allows one to process texts in various languages, to use it in information systems with natural-language interface, and to resolve such tasks as text classification and text search.

Recent technological developments have increased the use of machine learning to solve many problems, including many in information retrieval (IR). Deployment of machine-learning techniques is widespread in text search, notability web search engines (Dai et al., 2011). Multimedia information retrieval as a problem however still represents a significant challenge to machine learning as a technological solution, but some problems in IR can still be addressed by using appropriate AI techniques. In this paper we review the technological developments, and provide a perspective on the use of machine-learning techniques in conjunction with knowledge organisation techniques to address multimedia IR needs. We take the perspective from the MacFarlane (2016) position paper, that there are some problems in multimedia IR that AI and machine learning cannot currently solve. The semantic gap in multimedia IR (Enser, 2008) remains a significant problem in the field, and solutions to them are many years off. However, there are occasions where the new technological developments allow the use of knowledge organisation and machine learning in multimedia search systems and services. Specifically we argue that the improvement of detection of some classes of low level features in images (Karpathy and Li, 2015), music (Byrd and Crawford, 2002) and video (Hu et al., 2011) can be used in conjunction with knowledge organisation to tag or label multimedia content for better retrieval performance. We advocate the use of supervised learning techniques. We provide an overview of the use of knowledge organisation schemes in machine learning, and make recommendations to information professionals on the use of this technology with knowledge organisation techniques to solve multimedia IR problems.

SOMHunter represents a modern light-weight framework for known-item search in datasets of visual data like images or videos. The framework combines an effective W2VV++ text-to-image search approach, a traditional Bayesian like model for maintenance of relevance scores influenced by positive examples, and several types of exploration and exploitation displays. With this initial setting in 2020, already the first prototype of the system turned out to be highly competitive in comparison with other state-of-the-art systems at Video Browser Showdown and Lifelog Search Challenge competitions. In this paper, we present a new version of the system further extending the list of visual data search capabilities. The new version combines localized text queries with collage queries tested at VBS 2021 in two separate systems by our team. Furthermore, the new version of SOMHunter will integrate also the new CLIP text search model recently released by OpenAI. We believe that all the extensions will improve chances to effectively initialize the search that can continue with already supported browsing capabilities.

Natural Language Processing (NLP) field experienced considerable improvement in the year 2018. With the help of the transfer learning approach BERT (Bidirectional Encoder Representations) introduced by Google and approaches following it, state-of-the-art results were achieved in many NLP tasks. It became easier to research in this direction for Turkish with the help of published Turkish pre-trained BERT models in February, 2020. A new opportunity has arisen on research and development of contextual embeddings-based document retrieval instead of keyword-based for Turkish. This work introduces a BERT-contextual-vector-based Turkish text search tool. Sample queries and responses are presented to demonstrate the benefits it brings.

Information extraction from narrative clinical notes is useful for patient care, as well as for secondary use of medical data, for research or clinical purposes. Many studies focused on information extraction from English clinical texts, but less dealt with clinical notes in languages other than English. This study tested the feasibility of using 'off the shelf' information extraction algorithms to identify medical concepts from Italian clinical notes. We used MetaMap to map medical concepts to the Unified Medical Language System (UMLS). The study addressed two questions: (Q1) to understand if it would be possible to properly map medical terms found in clinical notes and related to the semantic group of 'Disorders' to the Italian UMLS resources; (Q2) to investigate if it would be feasible to use MetaMap as it is to extract these medical concepts from Italian clinical notes. Results in EXP1 showed that the Italian UMLS Metathesaurus sources covered 91% of the medical terms of the 'Disorders' semantic group, as found in the studied dataset. Even if MetaMap was built to analyze texts written in English, it worked properly also with texts written in Italian. MetaMap identified correctly about half of the concepts in the Italian clinical notes. Using MetaMap's annotation on Italian clinical notes instead of a simple text search improved our results of about 15 percentage points. MetaMap showed recall, precision and F-measure of 0.53, 0.98 and 0.69, respectively. Most of the failures were due to the impossibility for MetaMap to generate Italian meaningful variants. MetaMap's performance in annotating automatically translated English clinical notes was in line with findings in the literature, with similar recall (0.75), F-measure (0.83) and even higher precision (0.95)., This paper has been published in the Journal of biomedical informatics, Volume 63, October 2016, Pages 22-32

This paper presents an enhanced version of an interactive video retrieval tool SOMHunter that won Video Browser Showdown 2020. The presented enhancements focus on improving text querying capabilities since the text search model plays a crucial part in successful searches. Hence, we introduce the ability to specify multiple text queries with further positional specification so users can better describe positional relationships of the objects. Moreover, a possibility to further specify text queries with an example image is introduced as well as consequent changes to the user interface of the tool.

Keeping up with the rapid growth of Deep Learning (DL) research is a daunting task. While existing scientific literature search systems provide text search capabilities and can identify similar papers, gaining an in-depth understanding of a new approach or an application is much more complicated. Many publications leverage multiple modalities to convey their findings and spread their ideas - they include pseudocode, tables, images and diagrams in addition to text, and often make publicly accessible their implementations. It is important to be able to represent and query them as well. We utilize RDF Knowledge graphs (KGs) to represent multimodal information and enable expressive querying over modalities. In our demo we present an approach for extracting KGs from different modalities, namely text, architecture images and source code. We show how graph queries can be used to get insights into different facets (modalities) of a paper, and its associated code implementation. Our innovation lies in the multimodal nature of the KG we create. While our work is of direct interest to DL researchers and practitioners, our approaches can also be leveraged in other scientific domains.

As reported by respected evaluation campaigns focusing both on automated and interactive video search approaches, deep learning started to dominate the video retrieval area. However, the results are still not satisfactory for many types of search tasks focusing on high recall. To report on this challenging problem, we present two orthogonal task-based performance studies centered around the state-of-the-art W2VV++ query representation learning model for video retrieval. First, an ablation study is presented to investigate which components of the model are effective in two types of benchmark tasks focusing on high recall. Second, interactive search scenarios from the Video Browser Showdown are analyzed for two winning prototype systems implementing a selected variant of the model and providing additional querying and visualization components. The analysis of collected logs demonstrates that even with the state-of-the-art text search video retrieval model, it is still auspicious to integrate users into the search process for task types, where high recall is essential.

In the last decade, the Video Browser Showdown (VBS) became a comparative platform for various interactive video search tools competing in selected video retrieval tasks. However, the participation of new teams with an own, novel tool is prohibitively time-demanding because of the large number and complexity of components required for constructing a video search system from scratch. To partially alleviate this difficulty, we provide an open-source version of the lightweight known-item search system SOMHunter that competed successfully at VBS 2020. The system combines several features for text-based search initialization and browsing of large result sets; in particular a variant of W2VV++ model for text search, temporal queries for targeting sequences of frames, several types of displays including the eponymous self-organizing map view, and a feedback-based approach for maintaining the relevance scores inspired by PICHunter. The minimalistic, easily extensible implementation of SOMHunter should serve as a solid basis for constructing new search systems, thus facilitating easier exploration of new video retrieval ideas.

Raster-image PDF files originating from scanning or photographing paper documents are inaccessible to both text search engines and screen readers that people with visual impairments use. We here focus on the relatively less-researched problem of converting raster-image files with Arabic script into machine-accessible documents. Our method, called ECDP for "Ensemble-based classification of document patches," segments the physical layout of the document, classifies image patches as containing text or graphics, assembles homogeneous document regions, and passes the text to an optical character recognition engine to convert into natural language. Classification is based on the majority voting of an ensemble of support vector machines. When tested on the dataset BCE-Arabic [Saad et al. in: ACM 9th annual international conference on pervasive technologies related to assistive environments (PETRA'16), Corfu, 2016], ECDP yielded an average patch classification accuracy of 97.3% and average $$F_1$$F1 score of 95.26% for text patches and efficiently extracted text zones in both paragraphs and text-embedded graphics, even if the text is rotated by $$90^{\circ }$$90ź or is in English. ECDP outperforms a classical layout analysis method (RLSA) and a state-of-the-art commercial product (RDI-CleverPage) on this dataset and maintains a relatively high level of performance on document images drawn from two other datasets (Hesham et al. in Pattern Anal Appl 20:1275---1287, 2017; Proprietary Dataset of 109 Arabic Documents. http://www.rdi-eg.com). The results suggest that the proposed method has the potential to generalize well to the analysis of documents with a broad range of content.

More and more information on tourist spots is being represented as pictures rather than text. Consequently, tourists who are interested in a specific attraction shown in pictures may have no idea how to perform a text search to get more information about the interesting tourist spots. In the view of this problem and to enhance the competitiveness of the tourism market, this research proposes an innovative tourist spot identification mechanism, which is based on deep learning-based object detection technology, for real-time detection and identification of tourist spots by taking pictures on location or retrieving images from the Internet. This research establishes a tourist spot recognition system, which is a You Only Look Once version 3 model built in Tensorflow AI framework, and is used to identify tourist attractions by taking pictures with a smartphone's camera. To verify the possibility, a set of tourist spots in Hsinchu City, Taiwan is taken as an example. Currently, the tourist spot recognition system of this research can identify 28 tourist spots in Hsinchu. In addition to the attraction recognition feature, tourists can further use this tourist spot recognition system to obtain more information about 77 tourist spots from the Hsinchu City Government Information Open Data Platform, and then make dynamic travel itinerary planning and Google MAP navigation. Compared with other deep learning models using Faster region-convolutional neural networks or Single-Shot Multibox Detector algorithms for the same data set, the recognition time by the models using You Only Look Once version 3, Faster region-convolutional neural networks, and Single-Shot Multibox Detector algorithms are respectively 4.5, 5, and 9 s, and the mean average precision for each when IoU = 0.6 is 88.63%, 85%, and 43.19%, respectively. The performance experimental results of this research show the model using the You Only Look Once version 3 algorithm is more efficient and precise than the models using the Faster region-convolutional neural networks or the Single-Shot Multibox Detector algorithms, where You Only Look Once version 3 and Single-Shot Multibox Detector are one-stage learning architectures with efficient features, and Faster region-convolutional neural networks is a two-stage learning architecture with precise features.

We provide a quality evaluation of KB+Text search, a deep integration of knowledge base search and standard full-text search. A knowledge base (KB) is a set of subject–predicate–object triples with a common naming scheme. The standard query language is SPARQL, where queries are essentially lists of triples with variables. KB+Text search extends this by a special occurs-with predicate, which can be used to express the co-occurrence of words in the text with mentions of entities from the knowledge base. Both pure KB search and standard full-text search are included as special cases. We evaluate the result quality of KB+Text search on three different query sets. The corpus is the full version of the English Wikipedia (2.4 billion word occurrences) combined with the YAGO knowledge base (26 million triples). We provide a web application to reproduce our evaluation, which is accessible via http://ad.informatik.uni-freiburg.de/publications .

This work presents the bioMine system, a full-text natural language search engine for biomedical literature. bioMine provides search capabilities based on the full-text content of documents belonging to a database composed of scientific articles and allows users to submit their search queries using natural language. Beyond the text content of articles, the system engine also uses article metadata, empowering the search by considering extra information from picture and table captions. bioMine is publicly released as an open-source system under the MIT license.

Word stemming is a linguistic process in which the various inflected word forms are matched to their base form. It is among the basic text pre-processing approaches used in Natural Language Processing and Information Retrieval. Stemming is employed at the text pre-processing stage to solve the issue of vocabulary mismatch or to reduce the size of the word vocabulary, and consequently also the dimensionality of training data for statistical models. In this article, we present a fully unsupervised corpus-based text stemming method which clusters morphologically related words based on lexical knowledge. The proposed method performs cognitive-inspired computing to discover morphologically related words from the corpus without any human intervention or language-specific knowledge. The performance of the proposed method is evaluated in inflection removal (approximating lemmas) and Information Retrieval tasks. The retrieval experiments in four different languages using standard Text Retrieval Conference, Cross-Language Evaluation Forum, and Forum for Information Retrieval Evaluation collections show that the proposed stemming method performs significantly better than no stemming. In the case of highly inflectional languages, Marathi and Hungarian, the improvement in Mean Average Precision is nearly 50% as compared to unstemmed words. Moreover, the proposed unsupervised stemming method outperforms state-of-the-art strong language-independent and rule-based stemming methods in all the languages. Besides Information Retrieval, the proposed stemming method also performs significantly better in inflection removal experiments. The proposed unsupervised language-independent stemming method can be used as a multipurpose tool for various tasks such as the approximation of lemmas, improving retrieval performance or other Natural Language Processing applications.

This paper is intended to present the implementation and testing methodology of transductive support vector machines (TSVM) proposed by Joachims et al [1]. Initially it explains the concept offering by the Support Vector Machines as optimal classifiers and clarifies the concept of transductive inference. Along the implementation process several tests were performed. The data used for such tests was very diverse especially with respect to the dimensionality (number of samples, features,etc.). The ultimate objective was to integrate the Transductive inference tool in the already developed Intelligent Interface Web Engine [2] from the SISTA group at the Catholic University of Leuven (Belgium) [2].

While maintaining a food record is an essential means of health management, there has long been a reliance on conventional methods, such as entering text into record sheets, in the health medicine field. Food recording is a time-consuming activity; hence, there is a need for innovation using information technology. We have developed the smartphone application “FoodLog,” as a new framework for food recording. This application uses digital pictures and is supported by image recognition and searches. It is available for general release. In this paper, we present an overview of this framework, the data statistics obtained using FoodLog, and the future prospects of this application.

Leaflets are one of the most straightforward methods for obtaining details regarding how to administer drugs, their composition, warnings, or precautions in their usage. However, they are many times cluttered with information, making them hard to understand by people. At the same time, the provided indications are frequently either too specific or too broad, thus making their understanding even more difficult. This paper presents the architecture and the prototype of an intelligent platform for drug administration with the following key functionalities: adding specific medications to user’s profile, searching for potential contraindications or side effects, and defining alerts of drug administration in the provided calendar, based on the physician’s prescription. The platform aggregates leaflets from two sources (Biofarm and HelpNet) using Natural Language Processing (NLP) techniques tailored for Romanian language and provides full-text search via Elasticsearch. A friendly mobile application, available for IOS and Android devices, as well as in web browsers, was developed to improve users’ healthcare and ensure self-education.

ObjectiveElectronic health record (EHR) systems contain structured data and unstructured documentation. Clinical insights can be derived from analyzing both but optimal methods for this have not been studied extensively. We compared various approaches to analyzing EHR data for non-alcoholic fatty liver disease (NAFLD).Materials and MethodsWe compared analysis of structured and unstructured EHR data using natural language processing (NLP), free-text search, and diagnostic codes against expert adjudication as the reference standard.ResultsOut of 38,575 patients, we identified 2,281 patients with NAFLD. From the remainder, 10,653 patients with similar data density were selected as a control group. NLP was more sensitive than ICD and text search (NLP 0.93 vs. ICD 0.28 vs. text search 0.81) with higher a F2 score (NLP 0.92 vs. ICD 0.34 vs. text search 0.81). 619 patients had suspected NAFLD documented in radiology notes not acknowledged in other forms of clinical documentation. Of these, 232 (37.5%) were found to have more advanced liver disease after a median of 1,057 days.DiscussionNLP-based approaches have superior accuracy in identifying NAFLD within the EHR compared to ICD/text search-based approaches. Suspected NAFLD on imaging is often not acknowledged in subsequent clinical documentation. Many such patients are later found to have more advanced liver disease.ConclusionFor identification of NAFLD, NLP performed better than alternative selection modalities and facilitated follow-on analysis of information flow. If accuracy can be proven to persist across clinical domains, NLP can identify patient phenotypes for biomedical research in an accurate and high-throughput manner.

Since the 1970’s the Content-Based Image Indexing and Retrieval (CBIR) has been an active area. Nowadays, the rapid increase of video data has paved the way to the advancement of the technologies in many different communities for the creation of Content-Based Video Indexing and Retrieval (CBVIR). However, greater attention needs to be devoted to the development of effective tools for video search and browse. In this paper, we present Visione, a system for large-scale video retrieval. The system integrates several content-based analysis and retrieval modules, including a keywords search, a spatial object-based search, and a visual similarity search. From the tests carried out by users when they needed to find as many correct examples as possible, the similarity search proved to be the most promising option. Our implementation is based on state-of-the-art deep learning approaches for content analysis and leverages highly efficient indexing techniques to ensure scalability. Specifically, we encode all the visual and textual descriptors extracted from the videos into (surrogate) textual representations that are then efficiently indexed and searched using an off-the-shelf text search engine using similarity functions.

Design rationale (DR) is an important category within design knowledge, and effective reuse of it depends on its successful retrieval. In this paper, an ontology-based DR retrieval approach is presented, which allows users to search by entering normal queries such as questions in natural language. First, an ontology-based semantic model of DR is developed based on the extended issue-based information system-based DR representation in order to effectively utilize the semantics embedded in DR, and a database of ontology-based DR is constructed, which supports SPARQL queries. Second, two SPARQL query generation methods are proposed. The first method generates initial SPARQL queries from natural language queries automatically using template matching, and the other generates initial SPARQL queries automatically from DR record-based queries. In addition, keyword extension and optimization is conducted to enhance the SPARQL-based retrieval. Third, a design rationale retrieval prototype system is implemented. The experimental results show the advantages of the proposed approach.

The task of text/non-text classification in online handwritten documents is crucially important to text recognition, text search, and diagram interpretation. It, however, is a challenging problem because of the large amount of variation and lack of prior knowledge. In order to solve this problem, we propose to use global and local contexts to build a high-performance classifier. The classifier assigns a text or non-text label to each stroke in a stroke sequence of a digital ink document. First, a neural network architecture is used to acquire the complete global context of the sequence of strokes. Then, a simple but effective model based on a marginal distribution is used for the local temporal context of adjacent strokes in order to improve the sequence labeling result. The results of experiments on available heterogeneous online handwritten document databases demonstrate the superiority and effectiveness of our context combination approach. Our method achieved classification rates of 99.04% and 98.30% on the Kondate (written in Japanese) and IAMonDo (written in English) heterogeneous document databases. These results are significantly better than others reported in the literature. We present a method to combine global and local contexts for text/non-text classification in online handwritten documents.Global context refers to the feature vector sequence of an entire document and local context refers to the prediction of directly adjacent strokes.Global and local contexts are integrated by using a simple marginal distribution and basic combination rules.We propose multiple classifier combination strategies for combining global and local contexts to improve classification accuracy.

We propose a new compression algorithm that compresses plain texts by using a dictionary-based model and a compressed string-matching approach that can be used with the compressed texts produced by this algorithm. The compression algorithm (CAFTS) can reduce the size of the texts to approximately 41% of their original sizes. The presented compressed string matching approach (SoCAFTS), which can be used with any of the known pattern matching algorithms, is compared with a powerful compressed string matching algorithm (ETDC) and a compressed string-matching tool (Lzgrep). Although the search speed of ETDC is very good in short patterns, it can only search for exact words and its compression performance differs from one natural language to another because of its word-based structure. Our experimental results show that SoCAFTS is a good solution when it is necessary to search for long patterns in a compressed document.

This paper presents VISIONE, a tool for large-scale video search. The tool can be used for both known-item and ad-hoc video search tasks since it integrates several content-based analysis and re- trieval modules, including a keyword search, a spatial object-based search, and a visual similarity search. Our implementation is based on state-of- the-art deep learning approaches for the content analysis and leverages highly efficient indexing techniques to ensure scalability. Specifically, we encode all the visual and textual descriptors extracted from the videos into (surrogate) textual representations that are then efficiently indexed and searched using an off-the-shelf text search engine.

Sentiment analysis is a more popular area of highly active research in Automatic Language Processing. She assigns a negative or positive polarity to one or more entities using different natural language processing tools and also predicted high and low performance of various sentiment classifiers. Our approach focuses on the analysis of feelings resulting from reviews of products using original text search techniques. These reviews can be classified as having a positive or negative feeling based on certain aspects in relation to a query based on terms. In this paper, we chose to use two automatic learning methods for classification: Support Vector Machines (SVM) and Random Forest, and we introduce a novel hybrid approach to identify product reviews offered by Amazon. This is useful for consumers who want to research the sentiment of products before purchase, or companies that want to monitor the public sentiment of their brands. The results summarize that the proposed method outperforms these individual classifiers in this amazon dataset.

Translator is a tool to assist in translating a language from another language. Translator is very useful for tourists when traveling to other countries or other regions that do not understand the language of the country or area visited. Translator is also very useful for a student or student to learn and add knowledge about foreign languages that are not understood. The availability of translators in a university to improve services for students or those who need them. Students don't all understand foreign languages. String matching is a search technique in a text that is often called string search. String matching is often used in text search, location search, dictionary and so on. String matching has several algorithms, including brute force, boyer moore, knuth morris pratt and many more algorithms contained in string mathing. The Indonesian to English translator application uses the string mathing brute force algorithm. Brute force algorithm is matching each character from left to right.Keywords: Translator, String Matching, Algorithm, Brute Force

There are various approaches to storing numeric IDs as extra channels within CG rendered images. Using these channels, individual objects can be selected and separately modified. To associate an object with a text string a table or manifest is required mapping numeric IDs to text strings. This allows readable identification of ID-based selections, as well as the ability to make a selection using a text search. A scheme for storage of this ID Manifest is proposed which is independent of the approach used to store the IDs within the image. The total size of the raw strings within an ID Manifest may be very large but often contains much repeated information. A novel compression scheme is therefore employed which significantly reduces the size of the manifest.

Purely keyword-based text search is not satisfactory because named entities and WordNet words are also important elements to define the content of a document or a query in which they occur. Named entities have ontological features, namely, their aliases, classes, and identifiers. Words in WordNet also have ontological features, namely, their synonyms, hypernyms, hyponyms, and senses. Those features of concepts may be hidden from their textual appearance. Besides, there are related concepts that do not appear in a query, but can bring out the meaning of the query if they are added. We propose an ontology-based generalized Vector Space Model to semantic text search. It exploits ontological features of named entities and WordNet words, and develops a query-oriented spreading activation algorithm to expand queries. In addition, it combines and utilizes advantages of different ontologies for semantic annotation and searching. Experiments on a benchmark dataset show that, in terms of the MAP measure, our model is 42.5% better than the purely keyword-based model, and 32.3% and 15.9% respectively better than the ones using only WordNet or named entities. Keywords: semantic search, spreading activation, ontology, named entity, WordNet., 6 papes, Accepted by RIVF. arXiv admin note: substantial text overlap with arXiv:1807.05579; text overlap with arXiv:1807.05578

The main goal of existing word spotting approaches for searching document images has been the identification of visually similar word images in the absence of high quality text recognition output. Searching for a piece of arbitrary text is not possible unless the user identifies a sample word image from the document collection or generates the query word image synthetically. To address this problem, a Markov Random Field (MRF) framework is proposed for searching document images and shown to be effective for searching arbitrary text in real time for books printed in English (Latin script), Telugu and Ottoman scripts. The English experiments demonstrate that the dependencies between the visual terms and letter bigrams can be automatically learned using noisy OCR output. It is also shown that OCR text search accuracy can be significantly improved if it is combined with the proposed approach. No commercial OCR engine is available for Telugu or Ottoman script. In these cases the dependencies are trained using manually annotated document images. It is demonstrated that the trained model can be directly used to resolve arbitrary text queries across books despite font type and size differences. The proposed approach outperforms a state-of-the-art BLSTM baseline in these contexts.

Statistical stemmers are important components of Information Retrieval (IR) systems, especially for text search over languages with few linguistic resources. In recent years, research on stemmers produced relevant results, especially in 2011 when three language-independent stemmers were published in relevant venues. In this paper, we describe our efforts for reproducing these three stemmers. We also share the code as open-source and an extended version of Terrier system integrating the developed stemmers.

Major portion of web pages contains the natural language text and understanding of natural language text from the web pages is a major challenge for machines. Due to this lacking search engines are not able to provide relevant information to the users. This problem is tackled by natural language processing techniques and the development of ontologies from natural language text. With the help of such ontologies search of information can increases manifold. Much research in the field of text processing and automatic ontology building from text has been done to address these challenges. The proposed method in this paper is another effort to build automatic ontology from domain specific text. In this method we first extract concepts from a given domain specific text. We have used a Stanford parser to parse the text and the dictionary of basic concepts is created manually containing all the domain specific concepts and their relationships by recognizing laxico-syntactic patterns in the text corpus. Once concepts and relations among concepts as well as properties of concepts are identified, the extracted information can be represented in the form of graph and OWL.

Surnames play a key role as person natural identifiers, essentially in present information systems. This paper deals with the topic of optimizing a phonetic search algorithm as a string matching of surnames usable for communications service providers, person registries, social networks or genealogy databases. It describes a proposed solution for the phonetic searching of Slovak and (territorial) neighboring languages (Czech, Polish, Ukrainian, Russian, German, Hungarian, Jewish) surnames. This solution was designed to improve search precision and recall when searching for people by their surnames originating in these languages.

Numbers of UAV such as four or six rotor aircraft has been dramatically increased over recent years. And the volumetric surveillance based on UAV has been found to be extremely useful in haze visibility detection. however, owing to biunique flight control mode, the cost is higher and flexibility is limited, therefore, an effective human-UAV interaction is needed urgently, so that the UAV array could be controlled naturally by human hand. In this paper, a novel UAV gesture interaction scheme based on Kinect sensor and 2-D camera is presented. In the method, a kind of interaction interface is designed, then the UAV array is controlled in single or double hand. For gesture interaction details, based on the text search method proposed by Shahrouz, a new algorithm called “dummy independently component” is presented. In this approach, the finger movement is classified based on it's coherence, consequently, the number of parameters and database dimension are decreased, and search speed is improved. To verify the feasibility and effectiveness of scheme proposed above, final output is rendered with computer simulation and practical experiment.

Analysis of ancient Khmer documents can be quite challenging due to the elaborated shape of Khmer handwritten characters combined with the complex structure of how words are formed from those characters. Palm leaf manuscripts, one of the most well-known old Khmer documents, have been being digitized and centralized; therefore, document analysis functions such as text search capabilities are necessary but still remain unavailable for this type of documents. In order to contribute to the progress of relevant researches, we introduce in this paper a new dataset called SleukRith set comprising of 657 pages of Khmer palm leaf manuscripts randomly selected from various collections whose quality and digitization method are variable. The dataset contains three types of data: isolated characters, words, and lines. Each type of data is annotated with the ground truth information which is very useful for evaluating and serving as a training set for common document analysis tasks such as character/text recognition, word/line segmentation, and word spotting. In order to serve as a base line, the result of an evaluation study of Khmer isolated character recognition that we have conducted on SleukRith Set using Convolutional Neural Network is also presented.

In the bases of increasing the volume of text information, the dealing with text information has become incredibly complicated. The text clustering is a suitable technique used in dealing with a tremendous amount of text documents by classifying these set of text documents into clusters. Ultimately, text documents hold sparse, non-uniform distribution and uninformative features are difficult to cluster. The text feature selection is a primary unsupervised learning method that is utilized to choose a new subset of informational text features. In this paper, a new algorithm is proposed based on β-hill climbing technique for text feature selection problem to improve the text clustering (B-FSTC). The results of the proposed method for β-hill climbing and original Hill climbing (i.e., H-FSTC) are examined using the k-mean text clustering and compared with each other. Experiments were conducted on four standard text datasets with varying characteristics. Interestingly, the proposed β-hill climbing algorithm obtains superior results in comparison with the other well-regard techniques by producing a new subset of informational text features. Lastly, the β-hill climbing-based feature selection method supports the k-mean clustering algorithm to achieve more precise clusters.

Scene text localization and recognition keeps attracting an increasing interest from researchers due to its valuable advantage in extracting content from real world images and in image retrieval via text search. Nevertheless, due to the fact that the majority of the image datasets that are commonly used in this field is comprised of text in English, the related studies have mostly been limited to a single language. On that account, in order to apply the technologies developed for scene text detection and recognition to Turkish scene text, analyze their performances and to develop Turkish language specific algorithms, a Turkish scene text database is collected for the first time in the literature. In this paper, the contents of this database, shortly called STRIT (Scene Text Recognition In Turkish), are detailed. Additionally, two baseline methods are tested to detect and recognize scene text in Turkish and the preliminary results are presented.

A fundamental concern in search engines is to determine which documents have the best content for satisfying the user, based on analysis of the user's query and the text of documents. For this query-content match, many learning to rank systems make use of IR features developed in the 1990s in the TREC framework. Such features are still important in a variety of search tasks, and particularly in the long tail where clicks, links and social media signals become sparse. I will present our current progress, in particular three different neural models, with the goal of surpassing the 1990s models in full text search. This will include evidence, using proprietary Bing datasets, that large-scale training data can be useful. I will also argue that for the field to make progress on query-content relevance modeling, it may be valuable to set up a shared blind evaluation similar to 1990s TREC, possibly with large-scale training data.

In this paper, we adapt a surrogate text representation technique to develop efficient instance-level image retrieval using Regional Maximum Activations of Convolutions (R-MAC). R-MAC features have recently showed outstanding performance in visual instance retrieval. However, contrary to the activations of hidden layers adopting ReLU (Rectified Linear Unit), these features are dense. This constitutes an obstacle to the direct use of inverted indexes, which rely on sparsity of data. We propose the use of deep permutations, a recent approach for efficient evaluation of permutations, to generate surrogate text representation of R-MAC features, enabling indexing of visual features as text into a standard search-engine. The experiments, conducted on Lucene, show the effectiveness and efficiency of the proposed approach.

With the gradual increase of interest in plagiarism and intelligent file content search, the demand for similarity measuring between two sentences is increasing. There is a lot of researches for sentence similarity measurement methods in various directions such as n-gram, edit-distance and LSA. However, these methods have their own advantages and disadvantages. In this paper, we propose a new sentence similarity measurement method approaching from another direction. The proposed method uses the set-based POI data search that improves search performance compared to the existing hard matching method when data includes the inverse, omission, insertion and revision of characters. Using this method, we are able to measure the similarity between two sentences more accurately and more quickly. We modified the data loading and text search algorithm of the set-based POI data search. We also added a word operation algorithm and a similarity measure between two sentences expressed as a percentage. From the experimental results, we observe that our sentence similarity measurement method shows better performance than n-gram and the set-based POI data search.

A framework for fast text analysis, which is developed as a part of the Texterra project, is described. Texterra provides a scalable solution for the fast text processing on the basis of novel methods that exploit knowledge extracted from the Web and text documents. For the developed tools, details of the project, use cases, and evaluation results are presented.

 Abstract—In this paper a parallel optoelectronic computer architecture is proposed for large-scale parallel corpus, full text search and text mining applications while achieving high speed and high performance and utilizing the parallel processing nature of optics. An optical technique is presented that aims to automatically construct an Arabic-English corpus through web mining. In addition, it is shown how Arabic morphology and Arabic synonymous can be used to extract the required knowledge from Arabic texts based on converting the Arabic text into a semi-structured database. The proposed system is modeled and implemented using a two-dimensional space optics symbolic correlator to perform the comparison operations of the expert system. The proposed optical engine is able to perform 3D processing for inferring conclusions to queries by representing page contents in a 2D space using mapped-template logic. A hardware realization for these systems based on commercially available components is discussed. The performance analysis of the proposed system shows its remarkably improved processing performance with respect to processing speeds and both recall and precision measures.

While classical bag-of-word (BoG) approaches represent text content in the word level, recent studies show that knowledge-based concept indexation is a promising approach to further enhance the text search and mining performance. In this study, we propose a new knowledge indexation/extraction method, Explicit Semantic Path Mining (ESPM), for knowledge-base text mining. It has roots in a concept-based vector constructing method, Explicit Semantic Analysis (ESA), which has shown success in text mining tasks. For this new method, given an input piece of text, ESPM can efficiently identify the independent and optimized semantic path(s) on a concept map, which is, in this study, the Wikipedia category tree. Unlike earlier studies focusing on BoG based vector space, ESPM is a semantic path mining algorithm, which generates the top down semantic categories of a given text by leveraging the rich link information between Wikipedia categories and articles. Preliminary experiment based on ODP data shows ESPM delivers high quality independent semantic paths from both precision and ranking viewpoints.

This paper proposes a method for keyword spotting in off-line Chinese handwritten documents using a contextual word model, which measures the similarity between the query word and every candidate word in the document by combining a character classifier and the geometric context as well as linguistic context. The geometric context model characterizes the single-character likeliness and between-character relationship. The linguistic model utilizes the dependency of the word with the external adjacent characters. The combining weights are optimized on training documents. Experiments on a large handwriting database CASIA-HWDB demonstrate the effectiveness of the proposed method and justify the benefits of geometric and linguistic contexts. Compared to transcription-based text search, the proposed method can provide higher recall rate, and for spotting words of four characters, the proposed method provides both higher precision and recall rate.

The authors propose an online image search engine based on local image keypoint matching with GPU support. State-of-the-art models are based on bag-of-visual-words, which is an analogy of textual search for visual search. In this work, thanks to the vector computation power of the GPU, the authors utilize real values of keypoint descriptors and realize real-time search at keypoint level. By keeping the identities of each keypoint, closest keypoints are accurately retrieved. Image search has different characteristics than textual search. The authors implement one-to-one keypoint matching, which is more natural for images. The authors utilize GPUs for every basic step. To demonstrate practicality of GPU-extended image search, the authors also present a simple bag-of-visual-words search technique with full-text search engines. The authors explain how to implement one-to-one keypoint matching with text search engine. Proposed methods lead to drastic performance and precision improvement, which is demonstrated on datasets of different sizes.

We build an open-source toolkit which implements deterministic learning to support search and text classification tasks. We extend the mechanism of logical generalization towards syntactic parse trees and attempt to detect weak semantic signals from them. Generalization of syntactic parse tree as a syntactic similarity measure is defined as the set of maximum common sub-trees and performed at a level of paragraphs, sentences, phrases and individual words. We analyze semantic features of such similarity measure and compare it with semantics of traditional anti-unification of terms. Nearest-neighbor machine learning is then applied to relate a sentence to a semantic class. Using syntactic parse tree-based similarity measure instead of bag-of-words and keyword frequency approach, we expect to detect a weak semantic signal otherwise unobservable. The proposed approach is evaluated in a four distinct domains where a lack of semantic information makes classification of sentences rather difficult. We describe a toolkit which is a part of Apache Software Foun-dation project OpenNLP, designed to aid search engineers in tasks requiring text relevance assessment.