Comprehensive Syntax Research Articles

Fine-Grained Domain Adaptation for Chinese Syntactic Processing

Syntactic processing is fundamental to natural language processing. It provides rich and comprehensive syntax information in sentences that could be potentially beneficial for downstream tasks. Recently, pretrained language models have shown great success in Chinese syntactic processing, which typically involves word segmentation, POS tagging, and dependency parsing. However, the on-going research never ends since performance would be degraded drastically when tested on a highly-discrepant domain. This problem is widely accepted as domain adaptation, where the test domain differs from the training domain in supervised learning. Self-training is one promising solution for it, and straightforward source-to-target adaptation has already shown remarkable effectiveness in previous work. While this strategy ignores the fact that sentences of the target domain sentences may have very different gaps from the source training domain. More specifically, sentences with large gaps might fail by direct self-training adaptation. To this end, we propose fine-grained domain adaptation for Chinese syntactic processing in this work, aiming to model the gaps between the source and the target domains accurately and progressively. The key idea is to divide the target domain into fine-grained subdomains by using a specified domain distance metric, and then perform gradual self-training on the subdomains. We further offer an intuitive theoretical illustration based on the theory of Kumar et al. (2020) approximately. In addition, a novel representation learning framework is proposed to encode fine-grained subdomains effectively, aiming to utilize the above idea fully. Experimental results on benchmark datasets show that our method can achieve significant improvements over a variety of baselines.

Iterating on multiple collections in synchrony

Abstract Modern programming languages typically provide some form of comprehension syntax which renders programs manipulating collection types more readable and understandable. However, comprehension syntax corresponds to nested loops in general. There is no simple way of using it to express efficient general synchronized iterations on multiple ordered collections, such as linear-time algorithms for low-selectivity database joins. Synchrony fold is proposed here as a novel characterization of synchronized iteration. Central to this characterization is a monotonicisBeforepredicate for relating the orderings on the two collections being iterated on and an antimonotoniccanSeepredicate for identifying matching pairs in the two collections to synchronize and act on. A restriction is then placed on Synchrony fold, cutting its extensional expressive power to match that of comprehension syntax, giving us Synchrony generator. Synchrony generator retains sufficient intensional expressive power for expressing efficient synchronized iteration on ordered collections. In particular, it is proved to be a natural generalization of the database merge join algorithm, extending the latter to more general database joins. Finally, Synchrony iterator is derived from Synchrony generator as a novel form of iterator. While Synchrony iterator has the same extensional and intensional expressive power as Synchrony generator, the former is better dovetailed with comprehension syntax. Thereby, algorithms requiring synchronized iterations on multiple ordered collections, including those for efficient general database joins, become expressible naturally in comprehension syntax.

Representations and Optimizations for Embedded Parallel Dataflow Languages

Parallel dataflow engines such as Apache Hadoop, Apache Spark, and Apache Flink are an established alternative to relational databases for modern data analysis applications. A characteristic of these systems is a scalable programming model based on distributed collections and parallel transformations expressed by means of second-order functions such as map and reduce. Notable examples are Flink’s DataSet and Spark’s RDD programming abstractions. These programming models are realized as EDSLs—domain specific languages embedded in a general-purpose host language such as Java, Scala, or Python. This approach has several advantages over traditional external DSLs such as SQL or XQuery. First, syntactic constructs from the host language (e.g., anonymous functions syntax, value definitions, and fluent syntax via method chaining) can be reused in the EDSL. This eases the learning curve for developers already familiar with the host language. Second, it allows for seamless integration of library methods written in the host language via the function parameters passed to the parallel dataflow operators. This reduces the effort for developing analytics dataflows that go beyond pure SQL and require domain-specific logic. At the same time, however, state-of-the-art parallel dataflow EDSLs exhibit a number of shortcomings. First, one of the main advantages of an external DSL such as SQL—the high-level, declarative Select-From-Where syntax—is either lost completely or mimicked in a non-standard way. Second, execution aspects such as caching, join order, and partial aggregation have to be decided by the programmer. Optimizing them automatically is very difficult due to the limited program context available in the intermediate representation of the DSL. In this article, we argue that the limitations listed above are a side effect of the adopted type-based embedding approach. As a solution, we propose an alternative EDSL design based on quotations. We present a DSL embedded in Scala and discuss its compiler pipeline, intermediate representation, and some of the enabled optimizations. We promote the algebraic type of bags in union representation as a model for distributed collections and its associated structural recursion scheme and monad as a model for parallel collection processing. At the source code level, Scala’s comprehension syntax over a bag monad can be used to encode Select-From-Where expressions in a standard way. At the intermediate representation level, maintaining comprehensions as a first-class citizen can be used to simplify the design and implementation of holistic dataflow optimizations that accommodate for nesting and control-flow. The proposed DSL design therefore reconciles the benefits of embedded parallel dataflow DSLs with the declarativity and optimization potential of external DSLs like SQL.

Language Ability in the Intermediate-Scoring Group of Hearing-Impaired Children

Language development is a key issue in hearing-impaired children. However, interpersonal differences complicate our understanding of the situation. The bimodal or trimodal distribution of language scores in our other reports in this publication imply the presence of fundamental differences among these groups. The characteristic aspects of each group were profiled according to language data. We divided 268 children with prelingual severe to profound hearing impairment into 3 groups according to their trimodal distribution observed on histogram-based analysis of their responses to the Test of Question-Answer Interaction Development. Test results in several language domains, including productive and comprehensive vocabulary, productive and comprehensive syntax, and academic achievement, were profiled and compared among these 3 groups. Significant differences were observed in the results of the Word Fluency Test, the Picture Vocabulary Test-Revised, and the Syntax Test of Aphasia among the 3 groups. No significant difference was observed between groups who were lower-scoring and intermediate-scoring on the academic achievement tests referred to as Criterion Referenced Test-II and the Standardized Comprehension Test for Abstract Words. Only the higher-scoring group showed excellent results. The demographic factors were not significantly different among the 3 groups. Relatively poor academic achievement despite fair language production was the dominant feature of the intermediate-scoring group. This profile might correlate with academic failure in school.

The functional guts of the Kleisli query system

Kleisli is a modern data integration system that has made a significant impact on bioinformatics data integration. The primary query language provided by Kleisli is called CPL, which is a functional query language whose surface syntax is based on the comprehension syntax. Kleisli is itself implemented using the functional language SML. This paper describes the influence of functional programming research that benefits the Kleisli system, especially the less obvious ones at the implementation level.

Kleisli, a functional query system

Kleisli is a modern data integration system that has made a significant impact on bioinformatics data integration. This paper contains a brief introduction to the Kleisli system and an example to illustrate its uses in the bioinformatics arena. The primary query language provided by Kleisli is called CPL, which is a functional query language whose surface syntax is based on the comprehension syntax. Kleisli is itself implemented using the functional language SML. So this paper also describes the influence of functional programming research that benefits the Kleisli system, especially the less obvious ones at the implementation level.Availability. Kleisli has been commercialized under the name “KRIS”. It is available from Kris Technology Inc., 713 Santa Cruz Ave, #2, Menlo Park, CA 94025, USA. Direct email to info@kris-inc.com and web browser to http://www.kris-inc.com.

Comprehension syntax

The syntax of comprehensions is very close to the syntax of a number of practical database query languages and is, we believe, a better starting point than first-order logic for the development of database languages. We give an informal account of a language based on comprehension syntax that deals uniformly with a variety of collection types; it also includes pattern matching, variant types and function definition. We show, again informally, how comprehension syntax is a natural fragment of structural recursion, a much more powerful programming paradigm for collection types. We also show that a very small "abstract syntax language" can serve as a basis for the implementation and optimization of comprehension syntax.

Using English as a High-Level Robot Command Language

To facilitate expedient communication with robots, a very high-level hierarchical robot command language (Hirob) has been designed and implemented. Hirob uses the full and comprehensive syntax of the English imperative, allowing users to control a robot without the need of learning an esoteric programming language. A Parser/Scanner/Recognizer (PSR) performs a lexical analysis of a Hirob command stream, and identifies which portions of the command stream already exist as fully defined procedures in the files of the Procedure Management System (PMS). Those portions which do not exist must be defined using either existing Hirob procedures (English phrases), or by using the primitive commands of the low-level robot command language (Lorob). This process is fully recursive, so that Hirob procedures may consist of defined or undefined Hirob procedures, as well as Lorob commands, with the understanding that a high-level command cannot be executed until all of its hierarchical subcommands have been fully defined. A user-friendly editor has been incorporated into the PMS to allow convenient creation, modification, and testing of Hirob commands.

English as a command language for robot control (HIROB)

SUMMARYTo facilitate expedient communication with robots, a very-high level hierarchical robot command language (HIROB) has been designed and implemented. HIROB uses the full and comprehensive syntax of the English imperative, allowing users to control a robot without the need of learning an esoteric programming language. A Parser/Scanner/Recognizer (PSR) performs a lexical analysis of a HIROB command stream, and identifies which portions of the command stream already exist as fully defined procedures in the files of the Procedure Management System (PMS). Those portions which do not exist must be defined using either existing HIROB procedures (English phrases), or by using the primitive commands of the low-level robot command language (LOROB). This process is fully recursive, so that HIROB procedures may consist of defined or undefined HIROB procedures, as well as LOROB commands, with the understanding that a high-level command cannot be executed until all of its hierarchical sub-commands have been fully defined. A user-friendly editor has been incorporated into the PMS to allow convenient creation, modification, and testing of HIROB commands.