Parsing Speed Research Articles

DAT-MT Accelerated Graph Fusion Dependency Parsing Model for Small Samples in Professional Fields.

The rapid development of information technology has made the amount of information in massive texts far exceed human intuitive cognition, and dependency parsing can effectively deal with information overload. In the background of domain specialization, the migration and application of syntactic treebanks and the speed improvement in syntactic analysis models become the key to the efficiency of syntactic analysis. To realize domain migration of syntactic tree library and improve the speed of text parsing, this paper proposes a novel approach-the Double-Array Trie and Multi-threading (DAT-MT) accelerated graph fusion dependency parsing model. It effectively combines the specialized syntactic features from small-scale professional field corpus with the generalized syntactic features from large-scale news corpus, which improves the accuracy of syntactic relation recognition. Aiming at the problem of high space and time complexity brought by the graph fusion model, the DAT-MT method is proposed. It realizes the rapid mapping of massive Chinese character features to the model's prior parameters and the parallel processing of calculation, thereby improving the parsing speed. The experimental results show that the unlabeled attachment score (UAS) and the labeled attachment score (LAS) of the model are improved by 13.34% and 14.82% compared with the model with only the professional field corpus and improved by 3.14% and 3.40% compared with the model only with news corpus; both indicators are better than DDParser and LTP 4 methods based on deep learning. Additionally, the method in this paper achieves a speedup of about 3.7 times compared to the method with a red-black tree index and a single thread. Efficient and accurate syntactic analysis methods will benefit the real-time processing of massive texts in professional fields, such as multi-dimensional semantic correlation, professional feature extraction, and domain knowledge graph construction.

RabbitFX: Efficient Framework for FASTA/Q File Parsing on Modern Multi-Core Platforms.

The continuous growth of generated sequencing data leads to the development of a variety of associated bioinformatics tools. However, many of them are not able to fully exploit the resources of modern multi-core systems since they are bottlenecked by parsing files leading to slow execution times. This motivates the design of an efficient method for parsing sequencing data that can exploit the power of modern hardware, especially for modern CPUs with fast storage devices. We have developed RabbitFX, a fast, efficient, and easy-to-use framework for processing biological sequencing data on modern multi-core platforms. It can efficiently read FASTA and FASTQ files by combining a lightweight parsing method by means of an optimized formatting implementation. Furthermore, we provide user-friendly and modularized C++ APIs that can be easily integrated into applications in order to increase their file parsing speed. As proof-of-concept, we have integrated RabbitFX into three I/O-intensive applications: fastp, Ktrim, and Mash. Our evaluation shows that the inclusion of RabbitFX leads to speedups of at least 11.6 (6.6), 2.4 (2.4), and 3.7 (3.2) compared to the original versions on plain (gzip-compressed) files, respectively. These case studies demonstrate that RabbitFX can be easily integrated into a variety of NGS analysis tools to significantly reduce associated runtimes. It is open source software available at https://github.com/RabbitBio/RabbitFX.

Performance comparison of chosen JSON parsers and a parser that employs a different reading method

JSON parser, that has been developed, is made of two modules, each dedicated to process structure of either an object or an array. In both modules, building step of document model, that represents its structure, is combined with tokenization step. Design and development of algorithm is aimed at improving speed of parsing of documents.

Dependency Grammar Induction with a Neural Variational Transition-Based Parser

Dependency grammar induction is the task of learning dependency syntax without annotated training data. Traditional graph-based models with global inference achieve state-ofthe-art results on this task but they require O(n3) run time. Transition-based models enable faster inference with O(n) time complexity, but their performance still lags behind. In this work, we propose a neural transition-based parser for dependency grammar induction, whose inference procedure utilizes rich neural features with O(n) time complexity. We train the parser with an integration of variational inference, posterior regularization and variance reduction techniques. The resulting framework outperforms previous unsupervised transition-based dependency parsers and achieves performance comparable to graph-based models, both on the English Penn Treebank and on the Universal Dependency Treebank. In an empirical comparison, we show that our approach substantially increases parsing speed over graphbased models.

Deep Semantic Role Labeling With Self-Attention

Semantic Role Labeling (SRL) is believed to be a crucial step towards natural language understanding and has been widely studied. Recent years, end-to-end SRL with recurrent neural networks (RNN) has gained increasing attention. However, it remains a major challenge for RNNs to handle structural information and long range dependencies. In this paper, we present a simple and effective architecture for SRL which aims to address these problems. Our model is based on self-attention which can directly capture the relationships between two tokens regardless of their distance. Our single model achieves F1=83.4 on the CoNLL-2005 shared task dataset and F1=82.7 on the CoNLL-2012 shared task dataset, which outperforms the previous state-of-the-art results by 1.8 and 1.0 F1 score respectively. Besides, our model is computationally efficient, and the parsing speed is 50K tokens per second on a single Titan X GPU.

Fast and Efficient XML Data Access for Next-Generation Mass Spectrometry.

MotivationIn mass spectrometry-based proteomics, XML formats such as mzML and mzXML provide an open and standardized way to store and exchange the raw data (spectra and chromatograms) of mass spectrometric experiments. These file formats are being used by a multitude of open-source and cross-platform tools which allow the proteomics community to access algorithms in a vendor-independent fashion and perform transparent and reproducible data analysis. Recent improvements in mass spectrometry instrumentation have increased the data size produced in a single LC-MS/MS measurement and put substantial strain on open-source tools, particularly those that are not equipped to deal with XML data files that reach dozens of gigabytes in size.ResultsHere we present a fast and versatile parsing library for mass spectrometric XML formats available in C++ and Python, based on the mature OpenMS software framework. Our library implements an API for obtaining spectra and chromatograms under memory constraints using random access or sequential access functions, allowing users to process datasets that are much larger than system memory. For fast access to the raw data structures, small XML files can also be completely loaded into memory. In addition, we have improved the parsing speed of the core mzML module by over 4-fold (compared to OpenMS 1.11), making our library suitable for a wide variety of algorithms that need fast access to dozens of gigabytes of raw mass spectrometric data.AvailabilityOur C++ and Python implementations are available for the Linux, Mac, and Windows operating systems. All proposed modifications to the OpenMS code have been merged into the OpenMS mainline codebase and are available to the community at https://github.com/OpenMS/OpenMS.

A representational system of idiomatic constructions: For the building of computational resources

Chae, Hee-Rahk. 2014. A representational system of idiomatic constructions: For the building of computational resources. Linguistic Research 31(3), 491-518. We will provide a theory/platform-neutral framework for representing idiomatic expressions. Previous analyses on idioms, including those for some systems implemented on the basis of specific theories and platforms, are not flexible enough to deal with the gradational nature of syntactic and semantic anomalies of idioms, and/or are not explicit enough to be useful in natural language processing. We will overcome these problems by introducing some theory-neutral notational mechanisms, which apply to regular phrases, to capture the peculiarities of idioms. The present framework will become a very useful guide in building effective computational resources for a variety of languages, although we are focusing on the analysis of English and Korean idioms. One of the strong points of our approach is that we can formalize precisely the properties of potentially numerous different types of idioms, which is possible because we are using regular phrases built from phrase structure rules to represent the peculiarities of idiomatic expressions. The resulting resources will enhance the parsing accuracy and speed significantly, and will play an important role in disambiguation. (Hankuk University of Foreign Studies)

한국어 의존 관계 분석과 자질 집합 분할을 이용한 기계학습의 성능 개선

최근에 한국어 의존 관계에 대한 파싱 시스템과 관련된 연구가 소프트웨어 공학자들이나 언어학자들에 의해 다양하게 연구되고 있으며, 시스템 구현은 주로 기계 학습이나 기호 주의를 사용하고 있다. 기계 학습을 사용한 방법은 한국어 문장 데이터가 매우 크기 때문에 시스템 특성상 매우 긴 학습시간을 가지며, 데이터 자체가 가지는 오류로 인하여 한정된 인식율을 가진다. 본 연구에서는 기계학습을 이용한 시스템에 대하여 학습 시간을 줄일 수 있도록 특징들을 자질 집합 모듈로 분할하여 처리하는 방법을 제안하고, 문장수와 반복횟수에 따른 인식율을 분석하였다. 설계된 시스템은 분리된 모듈과 이진 검색을 위한 정렬 기법이 사용되었다. 데이터는 세종 말뭉치로부터 추출한 후 정제된 36,090문장을 사용하였다. 학습 시간은 약 3 시간으로 줄었으며, 인식율은 10,000 문장을 50회 학습하였을 때 84.54%로 가장 높았다. 모든 학습 문장(32,481)을 10회 학습하였을 때 인식율은 82.99%이다. 결과적으로 정제된 데이터를 이용하여 시스템이 안정화될 때까지 반복하는 것이 더 효율적이었다.

A Practical GLR Parser Generator for Software Reverse Engineering

Traditional parser generators use deterministic parsing methods. These methods can not meet the parsing requirements of software reverse engineering effectively. A new parser generator is presented which can generate GLR parser with automatic error recovery. The generated GLR parser has comparable parsing speed with the traditional LALR(1) parser and can be used in the parsing of software reverse engineering.

Distributed and Parallel Big Textual Data Parsing for Social Sensor Network

Recently, due to the popularization of the smartphone and social network service (SNS), many SNS users write their opinions for social events. According to these social phenomena, social sensor network which analyzes social events by utilizing those users' text data is proposed. Parsing is essential module to analyze user's text contents because it gives the understanding of semantics by extracting the words and their classes from texts. However, parsing requires much time because it needs to analyze all context information from the users' text. In addition, as users' text data are generated and transferred in streaming, the required parsing time increases too. This situation occurs that it is hard to parse the text on the single machine. Therefore, to drastically enhance the parsing speed, we propose distributed and parallel parsing system on the MapReduce. It applies the legacy parser to the MapReduce through loose coupling. Also, to reduce communication overheads, the statistical model used by the parser is resided on local cache in each mapper. The experimental result shows that the speed of proposed system is 2–19 times better than that of the legacy parser. As a result, we prove that the proposed system is useful for parsing text data in social sensor network.

Eye movements reset visual perception

Human vision uses saccadic eye movements to rapidly shift the sensitive foveal portion of our retina to objects of interest. For vision to function properly amidst these ballistic eye movements, a mechanism is needed to extract discrete percepts on each fixation from the continuous stream of neural activity that spans fixations. The speed of visual parsing is crucial because human behaviors ranging from reading to driving to sports rely on rapid visual analysis. We find that a brain signal associated with moving the eyes appears to play a role in resetting visual analysis on each fixation, a process that may aid in parsing the neural signal. We quantified the degree to which the perception of tilt is influenced by the tilt of a stimulus on a preceding fixation. Two key conditions were compared, one in which a saccade moved the eyes from one stimulus to the next and a second simulated saccade condition in which the stimuli moved in the same manner but the subjects did not move their eyes. We find that there is a brief period of time at the start of each fixation during which the tilt of the previous stimulus influences perception (in a direction opposite to the tilt aftereffect)--perception is not instantaneously reset when a fixation starts. Importantly, the results show that this perceptual bias is much greater, with nearly identical visual input, when saccades are simulated. This finding suggests that, in real-saccade conditions, some signal related to the eye movement may be involved in the reset phenomenon. While proprioceptive information from the extraocular muscles is conceivably a factor, the fast speed of the effect we observe suggests that a more likely mechanism is a corollary discharge signal associated with eye movement.

Development a New HL7 Interface Engine for Large-size Messages which Include Image Data based on Tree Structure and Streaming Algorithm

The almost existing commercial HL7 interface engines apply the string array method which is run in the main memory to HL7 message parsing process. But, if the HL7 message is big, this method will be possible to cause the computer system to raise critical and fatal problems because a long string array can carry a too heavy load to the main memory and the processor. Therefore, the image and the multi-media data which are needed for the modern medical records could be limited to be included into a HL7 message because the size is usually too big in comparison with the main body of a HL7 message and in result, it make the size of the HL7 message expanded. The purpose of this study is to suggest a new HL7 interface algorithm which can solve this problem by the method of the 'Streaming Algorithm'. This new method for HL7 message parsing apply the character-stream object which process character by character between the main memory and hard disk device with the consequence that the processing load on main memory could be alleviated. The main functions of this new engine are generating, parsing, validating, browsing, sending, and receiving of message. And also, this can parse and generate XML-formated HL7 message. This engine had been practically tested in the Discharge Summary Information Exchange System between Kyungpook National University Hospital and Chonnam National University Hospital for the purpose of proofing its usability for a month. Overall, the preliminary results of this test is considered as good, but it is pointed out that some improvement is needed relating to the speed of parsing which was predicted because this engine partly used the memory of hard disk device instead of the main memory. (Journal of Korean Society of Medical Informatics 10-1,17-33, 2004)

Efficient controlling of parsing-stack operation for LR parsers

This paper presents a new method of LR parsing based on the distinction of stack states and non-stack states. Non-stack states are states, which do not need to be pushed into the LR parsing stack, and stack states are states to be pushed into it. By using some properties based on the stack-controlling LR parser defined here, the parsing speed and the size of parsing tables can be remarkably improved and their improvement includes the traditional method eliminating unit productions. By empirical observations for a variety of programming languages, the efficiency is verified. Extending it to the generalized LR parsers for natural language is also discussed.

Even faster lr parsing

AbstractConventional LR parser generators create tables which are used to drive a standard parser procedure. Much faster parsers can be obtained by compiling the table entries into code that is directly executed. A possible drawback with a directly executable parser is its large size. In this paper, we introduce optimization techniques that increase the parsing speed even further while simultaneously reducing the size of the parser.

Very fast LR parsing

LR parsers can be made to run 6 to 10 times as fast as the best table-interpretive LR parsers. The resulting parse time is negligible compared to the time required by the remainder of a typical compiler containing the parser. A parsing speed of 1/2 million lines per minute on a computer similar to a VAX 11/780 was achieved, up from an interpretive speed of 40,000 lines per minute. A speed of 240,000 lines per minute on an Intel 80286 was achieved, up from an interpretive speed of 37,000 lines per minute. The improvement is obtained by translating the parser's finite state control into assembly language. States become code memory addresses. The current input symbol resides in a register and a quick sequence of register-constant comparisons determines the next state, which is merely jumped to. The parser's push-down stack is implemented directly on a hardware stack. The stack contains code memory addresses rather than the traditional state numbers. The strongly-connected components of the directed graph induced by the parser's terminal and nonterminal transitions are examined to determine a typically small subset of the states that require parse-time stack-overflow-check code when hardware does not provide the check automatically. The increase in speed is at the expense of space: a factor of 2 to 4 increase in total table size can be expected, depending upon whether syntactic error recovery is required.