Search

[Abstract]: Low-resource learning in natural language processing focuses on developing effective resources, tools, and technologies for languages that are less popular within the industry and academia. This effort is crucial for several reasons, including ensuring that as many languages as possible are represented digitally, and enhancing access to language technologies for native speakers of minority languages. In this context, this paper outlines the motivation, research lines, and results from a Leonardo Grant - by FBBVA - on low-resource languages and parsing as sequence labeling. The project’s primary aim was to devise fast and accurate methods for low-resource syntactic parsing and to examine evaluation strategies as well as strengths and weaknesses in comparison to alternative parsing strategies.

[Abstract]: Sentiment analysis is a key technology for companies and institutions to gauge public opinion on products, services or events. However, for large-scale sentiment analysis to be accessible to entities with modest computational resources, it needs to be performed in a resource-efficient way. While some efficient sentiment analysis systems exist, they tend to apply shallow heuristics, which do not take into account syntactic phenomena that can radically change sentiment. Conversely, alternatives that take syntax into account are computationally expensive. The SALSA project, funded by the European Research Council under a Proof-of-Concept Grant, aims to leverage recently-developed fast syntactic parsing techniques to build sentiment analysis systems that are lightweight and efficient, while still providing accuracy and explainability through the explicit use of syntax. We intend our approaches to be the backbone of a working product of interest for SMEs to use in production.

[Abstract]: Automatic grammar coaching serves an important purpose of advising on standard grammar varieties while not imposing social pressures or reinforcing established social roles. Such systems already exist but most of them are for English and few of them offer meaningful feedback. Furthermore, they typically rely completely on neural methods and require huge computational resources which most of the world cannot afford. We propose a grammar coaching system for Spanish that relies on (i) a rich linguistic formalism capable of giving informative feedback; and (ii) a faster parsing algorithm which makes using this formalism practical in a real-world application. The approach is feasible for any language for which there is a computerized grammar and is less reliant on expensive and environmentally costly neural methods. We seek to contribute to Greener AI and to address global education challenges by raising the standards of inclusivity and engagement in grammar coaching.

We generalize to a broader class of decoupled measures a result of Ziv and Merhav on universal estimation of the specific cross (or relative) entropy, originally for a pair of multi-level Markov measures. Our generalization focuses on abstract decoupling conditions and covers pairs of suitably regular g-measures and pairs of equilibrium measures arising from the “small space of interactions” in mathematical statistical mechanics.

A central topic in psycholinguistics is the study of how and when the parser assigns an antecedent to referentially-dependent elements. One such referentially-dependent element is the null subject of non-finite clauses. The aim of the present study was to examine the role of verb control information in the assignment of an antecedent to such a null subject. The results so far are inconclusive. Some authors argue that verb control information has a late influence, whereas others argue that such verb-specific information has a very rapid influence. We report a self-paced reading study in Spanish in which verb type (subject vs. object control) and grammaticality (grammatical vs. ungrammatical) were manipulated. The grammaticality manipulation was carried out by introducing a person anomaly at the infinitive itself, and not at a later word (e.g., "Te prometi/aconseje adelgazarme/adelgazarte cinco quilos en un mes." Literal translation, "I to you promised/advised to losemyself/yourself five kilos in a month"). With such a manipulation we can examine whether at the first possible point (i.e., the infinitive) verb control information was used to assign the correct antecedent (i.e., the subject in sentences with a subject-control verb, and the object in sentences with an object-control verb) to PRO. The results showed that at the infinitive there was a main effect of grammaticality, meaning that the correct antecedent has already been assigned to PRO. The present findings are consistent with models that assume that verb-specific information plays an important role in the initial stages of sentence processing.

Type systems (in the form of categorial grammars) are the front runners in the quest for a formally elegant, computationally attractive and flexible theory of linguistic form and meaning. Words enact typed constants, and interact with one another via means of grammatical rules enacted by type inferences, composing larger phrases in the process. The end result is at the same time a parse, a proof and a program, bridging the seemingly disparate fields of linguistics, formal logics and computer science. The transition from form to meaning is traditionally handled via a series of homomorphisms that simplify nuances of the syntactic calculus to move towards a uniform semantic calculus. Alluring as this might be, it poses pragmatic considerations. For the setup to work on the semantic level, one has no choice but to start from the hardest part, namely natural language syntax. Phenomena like movement, word-order variation, discontinuities, etc. require careful treatment that needs to be both general enough to encompass the full range of grammatical utterances, yet strict enough to ward off ungrammatical derivations. This thesis takes an operational shortcut in targeting a ``deeper'' calculus of grammatical composition, engaging only minimally with surface syntax. Where previously functional functional syntactic types would be position-conscious, requiring their arguments in predetermined positions upon a binary tree, they are now agnostic to both tree structure and sequential order, alleviating the need for syntactic refinements. This simplification comes at the cost of a misalignment between provability and grammaticality: the laxer semantic calculus permits more proofs than linguistically desired. To circumvent this underspecification, the thesis takes a step away from the established norm, proposing the incorporation of unary type operators extending the function-argument axis with grammatical role labels. The new calculus produces mixed unary/n-ary trees, each unary tr

A parallel algorithm for enumerating parse trees of a given string according to a fixed context-free grammar is defined. The algorithm computes the number of parse trees of an input string; more generally, it applies to computing the weight of a string in a weighted grammar. The algorithm is first implemented on an arithmetic circuit of depth O((log n)²) with O(n⁶) elements. Then, it is improved using fast matrix multiplication to use only O(n^5.38) elements, while preserving depth O((log n)²).

[Abstract] Dependency parsing is a crucial step towards deep language understanding and, therefore, widely demanded by numerous Natural Language Processing applications. In particular, left-to-right and top-down transition-based algorithms that rely on Pointer Networks are among the most accurate approaches for performing dependency parsing. Additionally, it has been observed for the top-down algorithm that Pointer Networks’ sequential decoding can be improved by implementing a hierarchical variant, more adequate to model dependency structures. Considering all this, we develop a bottom-up oriented Hierarchical Pointer Network for the left-to-right parser and propose two novel transition-based alternatives: an approach that parses a sentence in right-to-left order and a variant that does so from the outside in. We empirically test the proposed neural architecture with the different algorithms on a wide variety of languages, outperforming the original approach in practically all of them and setting new state-of-the-art results on the English and Chinese Penn Treebanks for non-contextualized and BERT-based embeddings.

[Abstract] In order to achieve deep natural language understanding, syntactic constituent parsing is a vital step, highly demanded by many artificial intelligence systems to process both text and speech. One of the most recent proposals is the use of standard sequence-to-sequence models to perform constituent parsing as a machine translation task, instead of applying task-specific parsers. While they show a competitive performance, these text-to-parse transducers are still lagging behind classic techniques in terms of accuracy, coverage and speed. To close the gap, we here extend the framework of sequence-to-sequence models for constituent parsing, not only by providing a more powerful neural architecture for improving their performance, but also by enlarging their coverage to handle the most complex syntactic phenomena: discontinuous structures. To that end, we design several novel linearizations that can fully produce discontinuities and, for the first time, we test a sequence-to-sequence model on the main discontinuous benchmarks, obtaining competitive results on par with task-specific discontinuous constituent parsers and achieving state-of-the-art scores on the (discontinuous) English Penn Treebank.

Semantics engineering tools like Redex can be used to define, explore, and debug formal definitions of programming language semantics. However, such tools are often based on abstract syntax, which makes the definition of rules and the exploration of execution traces rather unfriendly. In this paper we introduce Credex, a library in the Rascal meta-programming language for defining small-step evaluation-context semantics, where terms and matching patterns are what-you-see-is-what-you-get. Credex employs parsing for decomposing terms into context and redex. Since Rascal’s grammar formalism is based on general parsing, a non-unique decomposition of a term literally corresponds to an ambiguous parse. We demonstrate the use of Credex, detail some aspects of its implementation, and discuss three case-studies.

Some problems are considered solved by the research community. But are they really and does that mean we should stop investigating them? In this essay, I argue that "solved" problems often only appear solved on the surface, while fundamental open research problems lurk below the surface. It requires dedicated researchers to discover those open problems by applying the existing solutions and putting them to the test.

Semantics engineering tools like Redex can be used to define, explore, and debug formal definitions of programming language semantics. However, such tools are often based on abstract syntax, which makes the definition of rules and the exploration of execution traces rather unfriendly. In this paper we introduce Credex, a library in the Rascal meta-programming language for defining small-step evaluation-context semantics, where terms and matching patterns are what-you-see-is-what-you-get. Credex employs parsing for decomposing terms into context and redex. Since Rascal’s grammar formalism is based on general parsing, a non-unique decomposition of a term literally corresponds to an ambiguous parse. We demonstrate the use of Credex, detail some aspects of its implementation, and discuss three case-studies.

This paper discusses two approaches to parsing: Eelco Visser’s scannerless generalized LR parsing and our context-aware scanning paired with deterministic LR parsing. We compare the underlying techniques, specifically how parser context is used to disambiguate lexical syntax, and their use in the context of language evolution and composition applications. We also reflect on the many discussions shared with Eelco on these topics, and on our shared realization that our different assumptions about the contexts in which our approaches were used drove and justified the technical decisions made in each.

The specification of the syntax and semantics for Standard ML have been designed to support the generation of a compiler front end, but actual implementations have required significant modification to the specification. Since the specification was written there have been major advances in the development of language analysis systems that can handle general syntax specifications. We are revisiting the SML specification to consider to what extent, using modern tooling, it can be implemented exactly as originally written. In this short paper we focus on the lexical specification.

This essay surveys the current theoretical landscape of modern linguistics, asking whethe generative and possibly transformational linguistics can be applied to syntactic analysis of New Testament texts written in Koine Greek to find lingual hallmarks in the form of personal usage of syntactic rules that uniquely identify the authors of the texts. The conclusion is that there seems to be evidence that an application of a minimalist approach could make the detection of such lingual hallmarks possible.

Text classification is an essential task in natural language processing. While graph neural networks (GNNs) have successfully been applied to this problem both through graph classification and node classification approaches, their typical applications suffer from several issues. In the graph classification case, common graph construction techniques tend to leave out syntactic information. In the node classification case, most widespread datasets and applications tend to suffer from encoding relatively little information in the chosen node features. Finally, there are great benefits to be gained from combining the two GNN approaches. To tackle these concerns, we propose DepNet, a two-stage framework for text classification using GNN models. In the first stage we replace current graph construction methods by utilizing syntactic dependency parsing in order to include as much syntactic information in the GNN input as possible. In the second stage we combine both graph classification and node classification methods by utilizing the former to produce node embeddings for the latter, maximizing the potential of GNNs for text classification. We find that this technique significantly improves the performance of both graph classification and node classification approaches to text classification.

Several popular languages including Haskell and Python use the indentation and layout of code as an essential part of their syntax. In the past, implementations of these languages used ad hoc techniques to implement layout. Recent work has shown that a simple extension to context-free grammars can replace these ad hoc techniques and provide both formal foundations and efficient parsing algorithms for indentation sensitivity. However, that previous work is limited to bottom-up, LR(k) parsing, and many combinator-based parsing frameworks including Parsec use top-down algorithms that are outside its scope. This paper remedies this by showing how to add indentation sensitivity to parsing frameworks like Parsec. It explores both the formal semantics of and efficient algorithms for indentation sensitivity. It derives a Parsec-based library for indentation-sensitive parsing and presents benchmarks on a real-world language that show its efficiency and practicality. © 2014 ACM.

Several popular languages including Haskell and Python use the indentation and layout of code as an essential part of their syntax. In the past, implementations of these languages used ad hoc techniques to implement layout. Recent work has shown that a simple extension to context-free grammars can replace these ad hoc techniques and provide both formal foundations and efficient parsing algorithms for indentation sensitivity. However, that previous work is limited to bottom-up, LR(k) parsing, and many combinator-based parsing frameworks including Parsec use top-down algorithms that are outside its scope. This paper remedies this by showing how to add indentation sensitivity to parsing frameworks like Parsec. It explores both the formal semantics of and efficient algorithms for indentation sensitivity. It derives a Parsec-based library for indentation-sensitive parsing and presents benchmarks on a real-world language that show its efficiency and practicality. © 2014 ACM.

The front-end of a compiler reads the source program and performs analyses such as type checking. The goal of the front-end is to check for the presence of syntactic and semantic errors before the program is passed to the back-end of the compiler for tasks such as optimization and code generation. WebDSL is a domain-specific language for web programming that is being used for over 15 years. With WebDSL, many applications have been developed which have thousands of daily users. While the language has evolved over the years, the core of its implementation remains unchanged and is starting to show signs of a legacy system. The current WebDSL syntax is defined in SDF2 and all other parts of the compiler are implemented in Stratego. This thesis presents a modernized front-end of the WebDSL compiler, utilizing the meta-languages of the Spoofax language workbench. Specifically, we introduce a syntax definition of WebDSL in SDF3 that is implemented without the use of post-parse filters, and an executable declarative specification of the WebDSL static semantics in Statix. We use the modernized front-end as the largest case study to date for the meta-languages SDF3 and Statix, in order to evaluate their expressiveness, performance, and elegance when they are used to implement a real world language., Computer Science

Information exchange through the countless webservices is central in the current age of technology, which increases the importance for security in these developments. One such security feature is the validation of data based on standardized data structures. The aim of this thesis is to develop a flexible hardware-accelerated text-based recognizer that provides this strict syntax validation. To this end, a parsing machine architecture was adopted in order to fulfill the flexibility and strict recognition requirements. The parsing machine architecture was developed by formalizing the fundamental PEG expressions and creating a micro-architecture based on these PEG expressions, which led to the specification of the PPEG instruction set architecture. This architecture was then mathematically formalized and a proof for its strict adherence to the formalized PEG behavior was provided. The parsing machine architecture was implemented on an FPGA, a virtualization of the parsing machine was implemented in Python for easy analysis of its behavior, and a PEG compiler and assembler were developed for the PEG-PPEG translation. Finally, a memoization unit was developed as an extension to the parsing machine for an improved parsing throughput. By running benchmarks for CSV, XML, JSON, and Java files on the PPEG parsing machine implementation, its parsing behavior was analyzed and compared to existing solutions. This showed that the minimum stack sizes depend solely on the size and complexity of the PEG; the percentage of clock cycles spent on jumps in instruction and data memory is substantial, ranging from 18\% and 40\%; the PPEG-compiled binary code size is relatively small compared to other solutions; and the throughput of the PPEG parsing machine is comparable if not better than other solutions running on faster hardware. Finally, the memoization unit was found to benefit large complex grammars more than small grammars., Computer Engineering

[Abstract] Dependency and constituent trees are widely used by many artificial intelligence applications for representing the syntactic structure of human languages. Typically, these structures are separately produced by either dependency or constituent parsers. In this article, we propose a transition-based approach that, by training a single model, can efficiently parse any input sentence with both constituent and dependency trees, supporting both continuous/projective and discontinuous/non-projective syntactic structures. To that end, we develop a Pointer Network architecture with two separate task-specific decoders and a common encoder, and follow a multitask learning strategy to jointly train them. The resulting quadratic system, not only becomes the first parser that can jointly produce both unrestricted constituent and dependency trees from a single model, but also proves that both syntactic formalisms can benefit from each other during training, achieving state-of-the-art accuracies in several widely-used benchmarks such as the continuous English and Chinese Penn Treebanks, as well as the discontinuous German NEGRA and TIGER datasets.

It is well known that hyperedge-replacement grammars can generate NP-complete graph languages even under seemingly harsh restrictions. This means that the parsing problem is difficult even in the non-uniform setting, in which the grammar is considered to be fixed rather than being part of the input. Little is known about restrictions under which truly uniform polynomial parsing is possible. In this paper we propose a low-degree polynomial-time algorithm that solves the uniform parsing problem for a restricted type of hyperedge-replacement grammars which we expect to be of interest for practical applications.

Contract is a self-agreed, enforceable by law and deliberate agreement between two or more competent authority and parties. Contracts are made in written but may be implied or spoken, and generally have to do with another organization, employment, sale or lease, or tenancy. We assume service engagement is a part of business events. Business events such as payments, purchase, sells, delivery etc. not only impotent processes but are also inherently temporally constrained. Analysis phase is carried out to find out business event and their temporal relationships which helps business partners to analyze what to supply and what to require from others as its participates in the service engagement specified by a contract. Contracts are always be in unstructured text and their details also described in form of unstructured text. Our proposed system through this paper is to introduce a novel approach for employing classification, parsing to extract business event and their temporal constraints from contract text. Also we organize the event terms into cluster automatically with the use of topic modeling.

JSON is a popular data interchange format for many web, cloud, and IoT systems due to its simplicity, human readability, and widespread support. However, applications must first parse and convert the data to a native in-memory format before being able to perform useful computations. Many big data applications with high performance requirements convert JSON data to Apache Arrow RecordBatches, the latter being a widely-used columnar in-memory format for large tabular data sets used in data analytics. In this paper, we analyze the performance characteristics of such applications and show that JSON parsing represents a bottleneck in the system. Various strategies are explored to speed up JSON parsing on CPU and GPU as much as possible. Due to performance limitation of the CPU and GPU implementations, we furthermore present an FPGA accelerated implementation. We explain how hardware components that can parse variable-sized and nested structures can be combined to produce JSON parsers for any type of JSON document. Several fully integrated FPGA-accelerated JSON parser implementations are presented using the Intel Arria 10 GX and Xilinx VU37P devices, and compared to the performance of their respective host systems; an Intel Xeon and an IBM POWER9 system. Result show the accelerators achieve an end-to-end throughput close to 7 GB/s with the Arria 10 GX using PCIe, and close to 20 GB/s with the VU37P using OpenCAPI 3. Depending on the complexity of the JSON data to parse, the bandwidth is limited by the host-to-accelerator interface or available FPGA resources. Overall, this provides a throughput increase of up to 6x, compared to the baseline application. Also, we observe a full system energy efficiency improvement of up to 59x more JSON data parsed per joule., Computer Engineering

The Scannerless Generalized LR (SGLR) parsing algorithm supports the development of composed languages seamlessly but does not support incremental parsing. The Incremental Generalized LR (IGLR) parsing algorithm, on the other hand, does not support the seamless composition of languages. This thesis presents the Incremental Scannerless Generalized LR (ISGLR) parsing algorithm and investigates the effects of combining the SGLR and IGLR parsing algorithms. While the algorithmic differences are orthogonal, the fact that scannerless parsing relies on non-deterministic parsing for disambiguation has a negative impact on incrementality. Nonetheless, we show that the ISGLR parsing algorithm performs better than the batch SGLR parsing algorithm in typical scenarios. On average, the ISGLR parser can reuse 99% of a previous parse result. When parsing from scratch, the ISGLR parser has a 24% run time overhead compared to the SGLR parser, but when parsing incrementally for changes that are smaller than 1% of the input size on average, it has a 9× speedup., Successor of https://doi.org/10.1145/3359061.3361085, Computer Science

JSGLR2 is a modular Java implementation of the SGLR parsing algorithm that supports systematic benchmarking and improvement of its several parsing variants. By splitting the code into several components, they can be tested in isolation and thus optimized more effortlessly. The modular architecture, although beneficial for efficiently identifying and implementing optimizations, negatively impacts the performance of the parsing algorithm. This paper aims to measure the overhead introduced by the code architecture for one of the variants, more specifically the incremental variant, which combines incremental parsing with SGLR parsing. It does so by comparing the original implementation with a version with the modularity removed. The evaluation is done on programming languages used in practice: Java, WebDSL and SDF3. The results show that the inlined parser outperforms the previous one, achieving speedups of up to 16% in batch parsing and up to 10% in incremental parsing., CSE3000 Research Project, Computer Science and Engineering

JSGLR2 is a java implementation of the Scannerless Generalized LR-parsing (SGLR) algorithm. It employs a modular architecture. This architecture comes with a performance overhead for letting multiple components interact with each other. This paper looks into the size of the performance overhead penalty for the recovery parser variant. It does so by creating an 'inlined' version of the recovery parser variant. The inlined recovery variant is a JSGLR2 implementation that ignores the modular architecture, and hard-codes the components. The performance of the inlined variant is measured with a pre-existing evaluation suite. The results show that there is a performance increase between the original, and the inlined variant., CSE3000 Research Project, Computer Science and Engineering

This study investigates ‘fragments’ in contemporary English. Fragments are structurally noncanonical constituents that convey the propositional meaning of a full clause, such as Good Old Hendon next stop or What a weirdo. This investigation constitutes an innovative approach to the topic since it (i) explores fragments in exclusively written (i.e. planned/edited) discourse, and (ii) aims at providing a corpus-driven taxonomy and an empirical account of the constructions, strategies and phenomena that are classifiable as fragments based on linguistically objectifiable (formal/textual) criteria, two areas much neglected in prior literature. The results reveal that fragments are not uncommon in written registers, particularly in letters and novels/stories. The most frequent types identified are phrasal and verbless, followed by clausal, wh-fragments and Small Clauses. Most of them show a high rate of subject and/or verb omission whose recoverability in context is facilitated by means of functional elements or latent lexical items licensed by the construction itself., Este estudio investiga ‘fragmentos’ en inglés contemporáneo. Los fragmentos son constituyentes estructuralmente no canónicos que tienen el significado proposicional de una cláusula completa, como Good Old Hendon next stop o What a weirdo. Esta investigación constituye un enfoque innovador sobre el tema ya que (i) explora los fragmentos en el discurso escrito exclusivamente y (ii) tiene como objetivo elaborar una taxonomía basada en corpus y una descripción empírica de las construcciones, estrategias y fenómenos clasificables como fragmentos basada en criterios lingüísticos objetivables (formales/textuales), dos cuestiones poco exploradas en la literatura previa. Los resultados revelan que los fragmentos no son infrecuentes en los registros escritos, especialmente en las cartas y las novelas/historias. Los tipos más frecuentes identificados son frasales y sin verbo, seguidos de clausales, wh- y Small Clauses. La mayoría muestran una gran proporción de omisión de sujeto y/o verbo, recuperables en el contexto mediante elementos funcionales o elementos léxicos latentes justificados por la construcción en la que aparecen.

Contextual hyperedge replacement (CHR) strengthens the generative power of hyperedge replacement (HR) significantly, thus increasing its usefulness for practical modeling. We define top-down parsing for CHR grammars by graph transformation, and prove that it is correct as long as the generation and use of context nodes in productions does not create cyclic dependencies. An efficient predictive version of this algorithm can be obtained as in the case of HR grammars., Also part of the Theoretical Computer Science and General Issues book sub series (LNTCS, volume 12741)

Several popular languages including Haskell and Python use the indentation and layout of code as an essential part of their syntax. In the past, implementations of these languages used ad hoc techniques to implement layout. Recent work has shown that a simple extension to context-free grammars can replace these ad hoc techniques and provide both formal foundations and efficient parsing algorithms for indentation sensitivity. However, that previous work is limited to bottom-up, LR(k) parsing, and many combinator-based parsing frameworks including Parsec use top-down algorithms that are outside its scope. This paper remedies this by showing how to add indentation sensitivity to parsing frameworks like Parsec. It explores both the formal semantics of and efficient algorithms for indentation sensitivity. It derives a Parsec-based library for indentation-sensitive parsing and presents benchmarks on a real-world language that show its efficiency and practicality.

The Scannerless Generalized LR (SGLR) parsing algorithm supports the development of composed languages seamlessly but does not support incremental parsing. The Incremental Generalized LR (IGLR) parsing algorithm, on the other hand, does not support the seamless composition of languages. This thesis presents the Incremental Scannerless Generalized LR (ISGLR) parsing algorithm and investigates the effects of combining the SGLR and IGLR parsing algorithms. While the algorithmic differences are orthogonal, the fact that scannerless parsing relies on non-deterministic parsing for disambiguation has a negative impact on incrementality. Nonetheless, we show that the ISGLR parsing algorithm performs better than the batch SGLR parsing algorithm in typical scenarios. On average, the ISGLR parser can reuse 99% of a previous parse result. When parsing from scratch, the ISGLR parser has a 24% run time overhead compared to the SGLR parser, but when parsing incrementally for changes that are smaller than 1% of the input size on average, it has a 9× speedup., Successor of https://doi.org/10.1145/3359061.3361085, Computer Science

Existing evidence suggests that the parser avoids positing a movement dependency if the grammar does not require doing so. By investigating the processing of two syntactic ambiguities that have not been the subject of processing studies before, we provide more conclusive evidence for this parsing bias in two Romance languages: French and Italian. In two acceptability‐judgment experiments and two self‐paced‐reading studies, we found that sentences that involved a filler–gap dependency (indirect questions in Italian and free relatives in French) were dispreferred compared to sentences involving the same lexical material but no filler–gap dependency (declarative complement clauses in both languages). Crucially, the filler–gap dependency was not dispreferred when there was no available competitor. We conclude by discussing the relevance of these results for syntactic theory, in particular for the questionable status of Merge over Move as a grammatical principle.

During last years, researchers have focused on developing Deep Learning models to translate natural language questions to SQL structured queries, what is known as the "text-to-sql" translation task. By 2020, DL models have reached outstanding results in WikiSQL and Spider challenges. However, there are still some challenges they have not addressed. Time normalization and small size of training samples stand out among these challenges. Nowadays, state-of-the-art models for text-to-sql task are not able to deal with common questions like, "what were the average sales for last three Christmas?" or "how much did I sell in every weekend of the last quarter of previous year?". On the other hand, models are trained with relatively small number of samples compared to the number of parameters in the deep learning models and collecting more questions and annotate the dataset is not easy. For these reasons, we propose TimeSeq, a new standard for annotating temporal information as normalized token sequences (time sequences), which can take advantage of DL models that perform seq-to-seq translation for automatically normalizing relative time information contained in common questions about business databases. Currently this standard applies to the context of questions made about the content of a transactional database; but could grow to other domains. We demonstrated with our experiments that deep learning models can learn to summarize the temporal information contained in a natural language question into a time sequence. We developed a process for data augmentation to increase the examples of pairs of questions and time sequences, based in the substitution of paraphrases corresponding to 3 ontologies we integrated: (1) Ontology of Paraphrases for Temporal Expressions, (2) Ontology of Paraphrases for General Business Database Querying, and (3) Ontology of Paraphrases for Specific Industry Domain, and an algorithm to perform the data augmentation, by creating new pairs of question

This paper investigates the recent advances in parsing the Index Thomisticus Treebank, which encompasses Medieval Latin texts by Thomas Aquinas. The research focuses on two types of variables. On the one hand, it examines the impact that a larger dataset has on the results of parsing; on the other hand, performances of new parsers are analysed with respect to less recent tools. Term of comparison to determine the effective parsing advances are the results in parsing the Index Thomisticus Treebank described in a previous work. First, the best performing parser among those concerned in that study is tested on a larger dataset than the one originally used. Then, some parser combinations that were developed in the same study are evaluated as well, assessing that more training data result in more accurate performances. Finally, to examine the impact that newly available tools have on parsing results, we train, test, and evaluate two neural parsers chosen among those best performing in the CoNLL 2018 Shared Task. Our experiments reach the highest accuracy rates achieved so far in automatic syntactic parsing of the Index Thomisticus Treebank and of Latin overall.

[Abstract] Parsing is a core natural language processing technique that can be used to obtain the structure underlying sentences in human languages. Named entity recognition (NER) is the task of identifying the entities that appear in a text. NER is a challenging natural language processing task that is essential to extract knowledge from texts in multiple domains, ranging from financial to medical. It is intuitive that the structure of a text can be helpful to determine whether or not a certain portion of it is an entity and if so, to establish its concrete limits. However, parsing has been a relatively little-used technique in NER systems, since most of them have chosen to consider shallow approaches to deal with text. In this work, we study the characteristics of NER, a task that is far from being solved despite its long history; we analyze the latest advances in parsing that make its use advisable in NER settings; we review the different approaches to NER that make use of syntactic information; and we propose a new way of using parsing in NER based on casting parsing itself as a sequence labeling task.

Hyperedge replacement (HR) allows to define context-free graph languages, but parsing is NP-hard in the general case. Predictive top-down (PTD) is an efficient, backtrack-free parsing algorithm for subclasses of HR and contextual HR grammars, which has been described and implemented in earlier work, based on a representation of graphs and grammar productions as strings. In this paper, we define PTD parsers for HR grammars by graph transformation rules and prove that they are correct.