Paloma Moreda

Corpus-based semantic role approach in information retrieval

In this paper, a method to determine the semantic role for the constituents of a sentence is presented. This method, named SemRol, is a corpus-based approach that uses two different statistical models, conditional Maximum Entropy (ME) Probability Models and the TiMBL program, a Memory-based Learning. It consists of three phases that make use of features using words, lemmas, PoS tags and shallow parsing information. Our method introduces a new phase in the Semantic Role Labeling task which has usually been approached as a two phase procedure consisting of recognition and labeling arguments. From our point of view, firstly the sense of the verbs in the sentences must be disambiguated. That is why depending on the sense of the verb a different set of roles must be considered. Regarding the labeling arguments phase, a tuning procedure is presented. As a result of this procedure one of the best sets of features for the labeling arguments task is detected. With this set, that is different for TiMBL and ME, precisions of 76.71% for TiMBL or 70.55% for ME, are obtained. Furthermore, the semantic role information provided by our SemRol method could be used as an extension of Information Retrieval or Question Answering systems. We propose using this semantic information as an extension of an Information Retrieval system in order to reduce the number of documents or passages retrieved by the system.

Automatic Generalization of a QA Answer Extraction Module Based on Semantic Roles

In recent years, improvements on automatic semantic role labeling have grown the interest of researchers in its application to different NLP fields, specially to QA systems. We present a proposal of automatic generalization of the use of SR in QA systems to extract answers for different types of questions. Firstly, we have implemented two different versions of an answer extraction module using SR: a) rules-based, and b) patterns-based. These modules work as part of a QA system to extract answers for location questions. Secondly, these approaches have been automatically generalized to any type of factoid questions using generalization rules. The whole system has been evaluated using both location and temporal questions from TREC datasets. Results indicate that an automatic generalization is feasible, obtaining same quality results for both original type of questions and new auto-generalized one (Precision: 88.20% LOC and 95.08% TMP). Furthermore, results show that patterns-based approach works better in both types of questions (F1 improvement + 40.88% LOCand + 15.41% TMP).

The role of verb sense disambiguation in semantic role labeling

In this paper an exhaustive evaluation of the behavior of the most relevant features used in Semantic Role Disambiguation tasks when the senses of the verbs are considered and when they are not, is presented. This evaluation analyzes the influence of Verb Sense Disambiguation in the task. In order to do this, a whole system of Semantic Role Labeling is used and it is compared with similar methods. Our main results show how using the senses of the verbs improves the results for verb-specific roles, such as A2 or A3, and while not using them improves the results for adjuncts, such as modal or negative.

Text summarization contribution to semantic question answering: New approaches for finding answers on the web

As the Internet grows, it becomes essential to find efficient tools to deal with all the available information. Question answering (QA) and text summarization (TS) research fields focus on presenting the information requested by users in a more concise way. In this paper, the appropriateness and benefits of using summaries in semantic QA are analyzed. For this purpose, a combined approach where a TS component is integrated into a Web‐based semantic QA system is developed. The main goal of this paper is to determine to what extent TS can help semantic QA approaches, when using summaries instead of search engine snippets as the corpus for answering questions. In particular, three issues are analyzed: (i) the appropriateness of query‐focused (QF) summarization rather than generic summarization for the QA task, (ii) the suitable length comparing short and long summaries, and (iii) the benefits of using TS instead of snippets for finding the answers, tested within two semantic QA approaches (named entities and semantic roles). The results obtained show that QF summarization is better than generic (58% improvement), short summaries are better than long (6.3% improvement), and the use of TS within semantic QA improves the performance for both named‐entity‐based (10%) and, especially, semantic‐role‐based QA (47.5%).

TENOR: A Lexical Normalisation Tool for Spanish Web 2.0 Texts

The lexical richness and its ease of access to large volumes of information converts the Web 2.0 into an important resource for Natural Language Processing. Nevertheless, the frequent presence of non-normative linguistic phenomena that can make any automatic processing challenging. We therefore propose in this study the normalisation of non-normative lexical variants in Spanish Web 2.0 texts. We evaluate our system by restoring the canonical version of Twitter texts, increasing the F1 measure of a state-of-the-art approach for English texts by a 10%.

Identifying Semantic Roles Using Maximum Entropy Models

In this paper, a supervised learning method of semantic role labeling is presented. It is based on maximum entropy conditional probability models. This method acquires the linguistic knowledge from an annotated corpus and this knowledge is represented in the form of features. Several types of features have been analyzed for a few words selected from sections of the Wall Street Journal part of the Penn Treebank corpus.

MaNER: A MedicAl Named Entity Recogniser

This paper describes a medicinal products and active ingredients named entity recogniser (MaNER) for Spanish technical documents. This rule-based system uses high quality and low-maintenance lexicons. Our results (F-measure 90 %) proves that dictionary-based approaches, without any deep natural language processing (e.g. POS tagging), can achieve a high performance in this task. Our system obtains better results when compared to similar systems.

Two Proposals of a QA Answer Extraction Module Based on Semantic Roles

The contribution of semantic roles to question answering is considered to be very valuable. Due to this fact, the aim of this paper is to analyze the influence of semantic roles in this area. In order to achieve this goal a web QA system has been implemented using two different proposals for the answer extraction module based on semantic roles, and both implementations have been evaluated for location type questions. For the first proposal, a simple set of semantic rules was created, whereas, for the second proposal, a database of possible answer semantic patterns was automatically developed. This DB is created in a first step and it will be reused each time the answer extraction module is used. Results of both approaches have been analyzed and compared showing that the patterns-based approach improves the rules-based one in precision (+ 34.40%) and recall (+ 42.80%).

An Active Ingredients Entity Recogniser System Based on Profiles

This paper describes an active ingredients named entity recogniser. Our machine learning system, which is language and domain independent, employs unsupervised feature generation and weighting from the training data. The proposed automatic feature extraction process is based on generating a profile for the given entity without traditional knowledge resources (such as dictionaries). Our results (F1 87.3 % [95 %CI: 82.07–92.53]) proves that unsupervised feature generation can achieve a high performance for this task.

The study of informality as a framework for evaluating the normalisation of web 2.0 texts

The language used in Web 2.0 applications such as blogging platforms, realtime chats, social networks or collaborative encyclopaedias shows remarkable differences in comparison with traditional texts. The presence of informal features such as emoticons, spelling errors or Internet-specific slang can lower the performance of Natural Language Processing applications. In order to overcome this problem, text normalisation approaches can provide a clean word or sentence by transforming all non-standard lexical or syntactic variations into their canonical forms. Nevertheless, because the characteristics of each normalisation approach there exist different performance metrics and evaluation procedures. We hypothesize that the analysis of informality levels can be used to evaluate text normalization techniques. Thus, in this study we are going to propose a text normalisation evaluation framework using informality levels and its application to Web 2.0 texts.