Multi-Task Learning in Conditional Random Fields for Chunking in Shallow Semantic Parsing
1
Citation
15
Reference
20
Related Paper
Citation Trend
Abstract:
Alternating Structure Optimization (ASO) is a recently proposed linear Multitask Learning algorithm. Although its effective has been verified in both semi-supervised as well as supervised methods, yet they necessitate taking external resource as a prerequisite. Therefore, feasibility of employing ASO to further improve the performance merely rests on the labeled data on hand proves to be a task deserving close scrutiny. Catering to this challenging while untapped problem, this paper presents a novel application of ASO to the subtask of Shallow Semantic Parsing: Chunking. Our experiments on Chinese Treebank 5.0 present promising result in chunk analysis, and the error rate is reduced by 5.72%, proposing a profound way to further improve the performance.Keywords:
Treebank
Chunking (psychology)
Semantic role labeling
Word error rate
Supervised Learning
Cite
We formulate semantic parsing as a parsing problem on a synchronous context free grammar (SCFG) which is automatically built on the corpus of natural language sentences and the representation of semantic outputs. We then present an online learning framework for estimating the synchronous SCFG grammar. In addition, our online learning methods for semantic parsing problems are also extended to deal with the case, in which the semantic representation could be represented under lambda-calculus. Experimental results in the domain of semantic parsing show advantages in comparison with previous works.
Syntactic predicate
S-attributed grammar
Synchronous context-free grammar
Cite
Citations (0)
Traditional NLP has long held (supervised) syntactic parsing necessary for successful higher-level semantic language understanding (LU). The recent advent of end-to-end neural models, self-supervised via language modeling (LM), and their success on a wide range of LU tasks, however, questions this belief. In this work, we empirically investigate the usefulness of supervised parsing for semantic LU in the context of LM-pretrained transformer networks. Relying on the established fine-tuning paradigm, we first couple a pretrained transformer with a biaffine parsing head, aiming to infuse explicit syntactic knowledge from Universal Dependencies treebanks into the transformer. We then fine-tune the model for LU tasks and measure the effect of the intermediate parsing training (IPT) on downstream LU task performance. Results from both monolingual English and zero-shot language transfer experiments (with intermediate target-language parsing) show that explicit formalized syntax, injected into transformers through IPT, has very limited and inconsistent effect on downstream LU performance. Our results, coupled with our analysis of transformers' representation spaces before and after intermediate parsing, make a significant step towards providing answers to an essential question: how (un)availing is supervised parsing for high-level semantic natural language understanding in the era of large neural models?
S-attributed grammar
Cite
Citations (55)
Semantic parsing converts natural language queries into structured logical forms. The paucity of annotated training samples is a fundamental challenge in this field. In this work, we develop a semantic parsing framework with the dual learning algorithm, which enables a semantic parser to make full use of data (labeled and even unlabeled) through a dual-learning game. This game between a primal model (semantic parsing) and a dual model (logical form to query) forces them to regularize each other, and can achieve feedback signals from some prior-knowledge. By utilizing the prior-knowledge of logical form structures, we propose a novel reward signal at the surface and semantic levels which tends to generate complete and reasonable logical forms. Experimental results show that our approach achieves new state-of-the-art performance on ATIS dataset and gets competitive performance on Overnight dataset.
Logical form
Semantic role labeling
Cite
Citations (2)
This paper presents a novel method of semantic parsing that maps a natural language (NL) sentence to a logical form. We propose a semantic parsing method by conducting separately two steps as follows; 1) The first step is to predict semantic tags for a given input sentence. 2) The second step is to build a semantic representation structure for the sentence using the sequence of semantic tags. We formulate the problem of semantic tagging as a sequence learning using a conditional random field models (CRFs). We then represent a tree structure of a given sentence in which syntactic and semantic information are integrated in that tree. The learning problem is to map a given input sentence to a tree structure using a structure support vector model. Experimental results on the CLANG corpus show that the semantic tagging performance achieved a sufficiently high result. In addition, the precision and recall of mapping NL sentences to logical forms i.e. the meaning representation in CLANG show an improvement in comparison with the previous work.
CRFS
Semantic role labeling
Semantic compression
Explicit semantic analysis
Cite
Citations (0)
부분 의미 분석 시스템은 문장의 구성 요소들이 술어와 갖는 관계를 분석하는 것으로 문장에서 술어의 주체, 객체, 도구 등을 나타내는 의미 논항을 확인하게 된다. 본 논문에서 개발한 부분 의미 분석 시스템은 두 단계로 구성되어 있는데, 먼저 부분 구문 분석 결과로부터 의미 논항의 경계를 찾는 의미 논항 확인 단계를 수행하고 다음으로 확인된 의미 논항에 적절한 의미역을 부착하는 의미역 할당 단계를 수행한다. 순차적인 두 단계 방법을 적용하는 것에 의해서, 학습 성능 저하의 주요한 원인인 클래스 분포의 불균형 문제를 완화할 수 있고, 각 단계에 적합한 자질을 선별하여 사용할 수 있다. 본 논문에서는 PropBank 말뭉치에 기반한 CoNLL-2004 shared task의 데이터 집합 및 평가 프로그램을 사용하여 각 단계가 시스템의 전체 성능에 기여하는 정도를 보인다. A shallow semantic parsing system analyzes the relationship that a syntactic constituent of the sentence has with a predicate. It identifies semantic arguments representing agent, patient, instrument, etc. of the predicate. In this study, we propose a two-phase shallow semantic parsing model which consists of the identification phase and the classification phase. We first find the boundary of semantic arguments from partial syntactic parsing results, and then assign appropriate semantic roles to the identified semantic arguments. By taking the sequential two-phase approach, we can alleviate the unbalanced class distribution problem, and select the features appropriate for each task. Experiments show the relative contribution of each phase on the test data.
Predicate (mathematical logic)
Syntactic predicate
Semantic role labeling
Cite
Citations (1)
We present memory-based learning approaches to shallow parsing and apply these to five tasks: base noun phrase identification, arbitrary base phrase recognition, clause detection, noun phrase parsing and full parsing. We use feature selection techniques and system combination methods for improving the performance of the memory-based learner. Our approach is evaluated on standard data sets and the results are compared with that of other systems. This reveals that our approach works well for base phrase identification while its application towards recognizing embedded structures leaves some room for improvement.
Phrase
Identification
Determiner phrase
Feature (linguistics)
Cite
Citations (82)
The goal of semantic parsing is to map natural language to a machine interpretable meaning representation language (MRL). One of the constraints that limits full exploration of deep learning technologies for semantic parsing is the lack of sufficient annotation training data. In this paper, we propose using sequence-to-sequence in a multi-task setup for semantic parsing with focus on transfer learning. We explore three multi-task architectures for sequence-to-sequence model and compare their performance with the independently trained model. Our experiments show that the multi-task setup aids transfer learning from an auxiliary task with large labeled data to the target task with smaller labeled data. We see an absolute accuracy gain ranging from 1.0% to 4.4% in in our in-house data set and we also see good gains ranging from 2.5% to 7.0% on the ATIS semantic parsing tasks with syntactic and semantic auxiliary tasks.
Transfer of learning
Semantic role labeling
S-attributed grammar
Cite
Citations (74)
In this article, we study the problem of parsing a math problem into logical forms. It is an essential pre-processing step for automatically solving math problems. Most of the existing studies about semantic parsing mainly focused on the single-sentence level. However, for parsing math problems, we need to take the information of multiple sentences into consideration. To achieve the task, we formulate the task as a machine translation problem and extend the sequence-to-sequence model with a novel two-encoder architecture and a word-level selective mechanism. For training and evaluating the proposed method, we construct a large-scale dataset. Experimental results show that the proposed two-encoder architecture and word-level selective mechanism could bring significant improvement. The proposed method can achieve better performance than the state-of-the-art methods.
Sequence (biology)
Cite
Citations (5)
Supervised training procedures for semantic parsers produce high-quality semantic parsers, but they have difficulty scaling to large databases because of the sheer number of logical constants for which they must see labeled training data. We present a technique for developing semantic parsers for large databases based on a reduction to standard supervised training algorithms, schema matching, and pattern learning. Leveraging techniques from each of these areas, we develop a semantic parser for Freebase that is capable of parsing questions with an F1 that improves by 0.42 over a purely-supervised learning algorithm.
Semantic Matching
Schema Matching
Schema (genetic algorithms)
Cite
Citations (279)
Chunking or shallow syntactic parsing is proving to be a task of interest to many natural language processing applications. The problem gets worse for the Arabic language because of its specific features that make it quite different and even more ambiguous than other natural languages when processed. In this paper, we present a method for chunking Arabic texts based on supervised learning. We use the Conditional Random Fields algorithm and the Penn Arabic Treebank to train the model. For the experimentation, we use over than 10,100 sentences as training data and 2,524 sentences for the test. The evaluation of the method consists of the calculation of the generated model accuracy and the results are very encouraging.
Treebank
Chunking (psychology)
Cite
Citations (6)