Information For Disambiguation Research Articles

A semantic union model for open domain Chinese knowledge base question answering

In Open-domain Chinese Knowledge Base Question Answering (ODCKBQA), most common simple questions can be answered by a single relational fact in the knowledge base (KB). The abbreviations, aliases, and nesting of entities in Chinese question sentences, and the gap between them and the structured semantics in the knowledge base, make it difficult for the system to accurately return answers. This study proposes a semantic union model (SUM), which concatenates candidate entities and candidate relationships, using a contrastive learning algorithm to learn the semantic vector representation of question and candidate entity-relation pairs, and perform cosine similarity calculations to simultaneously complete entity disambiguation and relation matching tasks. It can provide information for entity disambiguation through the relationships between entities, avoid error propagation, and improve the system performance. The experimental results show that the system achieves a good average F1 of 85.94% on the dataset provided by the NLPCC-ICCPOL 2016 KBQA task.

Learning about the meanings of ambiguous words: evidence from a word-meaning priming paradigm with short narratives.

Fluent language comprehension requires people to rapidly activate and integrate context-appropriate word meanings. This process is challenging for meanings of ambiguous words that are comparatively lower in frequency (e.g., the “bird” meaning of “crane”). Priming experiments have shown that recent experience makes such subordinate (less frequent) word meanings more readily available at the next encounter. These experiments used lists of unconnected sentences in which each ambiguity was disambiguated locally by neighbouring words. In natural language, however, disambiguation may occur via more distant contextual cues, embedded in longer, connected communicative contexts. In the present experiment, participants (N = 51) listened to 3-sentence narratives that ended in an ambiguous prime. Cues to disambiguation were relatively distant from the prime; the first sentence of each narrative established a situational context congruent with the subordinate meaning of the prime, but the remainder of the narrative did not provide disambiguating information. Following a short delay, primed subordinate meanings were more readily available (compared with an unprimed control), as assessed by responses in a word association task related to the primed meaning. This work confirms that listeners reliably disambiguate spoken ambiguous words on the basis of cues from wider narrative contexts, and that they retain information about the outcome of these disambiguation processes to inform subsequent encounters of the same word form.

3D Model Classification Based on GCN and SVM

3D model classification is an important task. Now, 3D model is usually expressed as point cloud. Disorder of point cloud brings great difficulty into 3D model classification. In order to classify 3D model correctly, a new classification method combining Graph Convolution Network (GCN) and Support Vector Machine (SVM) is proposed in this paper. Point cloud is sampled. K-Nearest Neighbor ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">KNN</i> ) algorithm is used to find K nearest points of sampling point, and adjacency matrix is established for graph convolution operation. Shape features <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">D1, D2, D3</i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">A3</i> of sampling point are computed based on its K nearest points. Coordinates and shape features of sampling point are combined as discriminative feature. 2-layer graph convolution is used to aggregate disambiguation information of 1-degree and 2-degree adjacent points of sampling point for describing point cloud comprehensively. At the same time, maximum pooling and average pooling are adopted to retain representative information. Finally, SVM is used to classify point clouds. Experimental results show that compared with GCN based on coordinates, the proposed network improves accuracy of 3D model classification by 1.67%. Global and local information can be extracted adequately when 1024 points are sampled from point cloud. When we select 20 nearest points to compute shape features <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">D1, D2, D3, A3</i> , local information of point can be described better. Shape features <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">D1, D2, D3, A3</i> are combined with coordinates to describe shape and structure of point cloud better. 2-layer graph convolutions are adopted to aggregate information of 1-degree and 2-degree nodes for extracting effective disambiguation features.

Priming syntactic ambiguity resolution in children and adults

ABSTRACT Adults use their recent experience to disambiguate ambiguous sentences: Structures that have recently been primed are favoured in the resolution of different types of ambiguity, an example of structural priming. Research on children's use of recent information for disambiguation is scarce. Using a forced-choice task with a tablet, we asked whether 5–6-year-old French-speaking children could also be primed in the resolution of attachment ambiguities, as well as whether listeners are affected by the proportion of primes of each structure, and whether priming is cumulative. We found that both children and adults can be primed, and are sensitive to the proportion of structures in the input, and that priming effects cumulate as the experiment progresses. This is the first study showing priming of ambiguous sentences at 5–6 years, suggesting that children, like adults, use recent experience as a source of disambiguating information.

Greater Early Disambiguating Information for Less-Probable Words: The Lexicon Is Shaped by Incremental Processing.

There has been much work over the last century on optimization of the lexicon for efficient communication, with a particular focus on the form of words as an evolving balance between production ease and communicative accuracy. Zipf’s law of abbreviation, the cross-linguistic trend for less-probable words to be longer, represents some of the strongest evidence the lexicon is shaped by a pressure for communicative efficiency. However, the various sounds that make up words do not all contribute the same amount of disambiguating information to a listener. Rather, the information a sound contributes depends in part on what specific lexical competitors exist in the lexicon. In addition, because the speech stream is perceived incrementally, early sounds in a word contribute on average more information than later sounds. Using a dataset of diverse languages, we demonstrate that, above and beyond containing more sounds, less-probable words contain sounds that convey more disambiguating information overall. We show further that this pattern tends to be strongest at word-beginnings, where sounds can contribute the most information.

Using transitivity information for morphological and syntactic disambiguation of pronouns in Ukrainian

Using transitivity information for morphological and syntactic disambiguation of pronouns in Ukrainian

Applications of Non-Uniquely Decodable Codes to Privacy-Preserving High-Entropy Data Representation

Non-uniquely-decodable (non-UD) codes can be defined as the codes that cannot be uniquely decoded without additional disambiguation information. These are mainly the class of non–prefix–free codes, where a code-word can be a prefix of other(s), and thus, the code-word boundary information is essential for correct decoding. Due to their inherent unique decodability problem, such non-UD codes have not received much attention except a few studies, in which using compressed data structures to represent the disambiguation information efficiently had been previously proposed. It had been shown before that the compression ratio can get quite close to Huffman/Arithmetic codes with an additional capability of providing direct access in compressed data, which is a missing feature in the regular Huffman codes. In this study we investigate non-UD codes in another dimension addressing the privacy of the high-entropy data. We particularly focus on such massive volumes, where typical examples are encoded video or similar multimedia files. Representation of such a volume with non–UD coding creates two elements as the disambiguation information and the payload, where decoding the original data from these elements becomes hard when one of them is missing. We make use of this observation for privacy concerns. and study the space consumption as well as the hardness of that decoding. We conclude that non-uniquely-decodable codes can be an alternative to selective encryption schemes that aim to secure only part of the data when data is huge. We provide a freely available software implementation of the proposed scheme as well.

Неоднозначность в лингвистике и переводе

Ambiguity implies that there are at least two distinct senses ascribed to one sign. It is in­herent to language and speech. In this article, I reflect on the types of ambiguity, its typology, production and effect and propose an algorithm for tackling ambiguity in translation. I posit that the choice of a translation strategy and the need for disambiguation in general depend on the type of ambiguity, its sources and character, i. e. whether ambiguity is intended or not. Intended ambiguity occurs when the speaker intentionally does not follow the logic of concep­tual clues (primes) and opts for a set of communicative strategies and linguistic means, which allow him/her to offer several possible interpretations of one event or even refer to several dif­ferent events. I explore a rarely analyzed event-referential ambiguity, which requires addi­tional conceptual information for disambiguation and, consequently, may pose a problem for translation. I argue that problems in disambiguation may occur for a variety of reasons: the translator and\or the recipient may have a wrong reference, have insufficient background knowledge to resolve the ambiguity or make wrong inferences since each recipient bears a different combination of cognitive, axiological, social, professional and gender attributes.

Combining range and inequality information for pointer disambiguation

Pentagons is an abstract domain invented by Logozzo and Fähndrich to validate array accesses in low-level programming languages. This algebraic structure provides a cheap “less-than check”, which builds a partial order between the integer variables used in a program. In this paper, we show how we have used the ideas available in Pentagons to design and implement a novel alias analysis. With this new algorithm, we are able to disambiguate pointers with offsets, that commonly occur in C programs, in a precise and efficient way. Together with this new abstract domain we describe several implementation decisions that let us produce a practical pointer disambiguation algorithm on top of the LLVM compiler. Our alias analysis is able to handle programs as large as SPEC CPU2006's gcc in a few minutes. Furthermore, it improves on LLVM's industrial quality analyses. As an extreme example, we have observed a 4x improvement when analyzing SPEC's lbm .

Adaptive plasticity in speech perception: Effects of external information and internal predictions.

When listeners encounter speech under adverse listening conditions, adaptive adjustments in perception can improve comprehension over time. In some cases, these adaptive changes require the presence of external information that disambiguates the distorted speech signals, whereas in other cases mere exposure is sufficient. Both external (e.g., written feedback) and internal (e.g., prior word knowledge) sources of information can be used to generate predictions about the correct mapping of a distorted speech signal. We hypothesize that these predictions provide a basis for determining the discrepancy between the expected and actual speech signal that can be used to guide adaptive changes in perception. This study provides the first empirical investigation that manipulates external and internal factors through (a) the availability of explicit external disambiguating information via the presence or absence of postresponse orthographic information paired with a repetition of the degraded stimulus, and (b) the accuracy of internally generated predictions; an acoustic distortion is introduced either abruptly or incrementally. The results demonstrate that the impact of external information on adaptive plasticity is contingent upon whether the intelligibility of the stimuli permits accurate internally generated predictions during exposure. External information sources enhance adaptive plasticity only when input signals are severely degraded and cannot reliably access internal predictions. This is consistent with a computational framework for adaptive plasticity in which error-driven supervised learning relies on the ability to compute sensory prediction error signals from both internal and external sources of information. (PsycINFO Database Record

Predicting the unbeaten path through syntactic priming.

A number of previous studies showed that comprehenders make use of lexically based constraints such as subcategorization frequency in processing structurally ambiguous sentences. One piece of such evidence is lexically specific syntactic priming in comprehension; following the costly processing of a temporarily ambiguous sentence, comprehenders experience less processing difficulty with the same structure with the same verb in subsequent processing. In previous studies using a reading paradigm, however, the effect was observed at or following disambiguating information and it is not known whether a priming effect affects only the process of resolving structural ambiguity following disambiguating input or it also affects the process before ambiguity is resolved. Using a visual world paradigm, the current study addressed this issue with Japanese relative clause sentences. Our results demonstrated that after experiencing the relative clause structure, comprehenders were more likely to predict the usually dispreferred structure immediately upon hearing the same verb. No compatible effect, in contrast, was observed on hearing a different verb. Our results are consistent with the constraint-based lexicalist view, which assumes the parallel activation of possible structural analyses at the verb. Our study demonstrated that an experience of a dispreferred structure activates the structural information in a lexically specific manner, leading comprehenders to predict another instance of the same structure on encountering the same verb.

Immediate use of prosody and context in predicting a syntactic structure

Numerous studies have reported an effect of prosodic information on parsing but whether prosody can impact even the initial parsing decision is still not evident. In a visual world eye-tracking experiment, we investigated the influence of contrastive intonation and visual context on processing temporarily ambiguous relative clause sentences in Japanese. Our results showed that listeners used the prosodic cue to make a structural prediction before hearing disambiguating information. Importantly, the effect was limited to cases where the visual scene provided an appropriate context for the prosodic cue, thus eliminating the explanation that listeners have simply associated marked prosodic information with a less frequent structure. Furthermore, the influence of the prosodic information was also evident following disambiguating information, in a way that reflected the initial analysis. The current study demonstrates that prosody, when provided with an appropriate context, influences the initial syntactic analysis and also the subsequent cost at disambiguating information. The results also provide first evidence for pre-head structural prediction driven by prosodic and contextual information with a head-final construction.

A Unified Probabilistic Framework for Name Disambiguation in Digital Library

Despite years of research, the name ambiguity problem remains largely unresolved. Outstanding issues include how to capture all information for name disambiguation in a unified approach, and how to determine the number of people K in the disambiguation process. In this paper, we formalize the problem in a unified probabilistic framework, which incorporates both attributes and relationships. Specifically, we define a disambiguation objective function for the problem and propose a two-step parameter estimation algorithm. We also investigate a dynamic approach for estimating the number of people K. Experiments show that our proposed framework significantly outperforms four baseline methods of using clustering algorithms and two other previous methods. Experiments also indicate that the number K automatically found by our method is close to the actual number.

Topological map induction using neighbourhood information of places

In topological mapping, perceptual aliasing can cause different places to appear indistinguishable to the robot. In case of severely corrupted or non-available odometry information, topological mapping is difficult as the robot is challenged with the loop-closing problem; that is to determine whether it has visited a particular place before. In this article we propose to use neighbourhood information to disambiguate otherwise indistinguishable places. Using neighbourhood information for place disambiguation is an approach that neither depends on a specific choice of sensors nor requires geometric information such as odometry. Local neighbourhood information is extracted from a sequence of observations of visited places. In experiments using either sonar or visual observations from an indoor environment the benefits of using neighbourhood clues for the disambiguation of otherwise identical vertices are demonstrated. Over 90% of the maps we obtain are isomorphic with the ground truth. The choice of the robot's sensors does not impact the results of the experiments much.

Impaired Evidence Integration and Delusions in Schizophrenia

A bias against disconfirmatory evidence (BADE) appears to be related to delusions in schizophrenia. However, preliminary studies have either not used the most comprehensive version of the BADE task, not included a psychiatric control group, and/or have used difference score methodology instead of analyzing all available measures. In the current study a comprehensive version of the BADE task was administered to people with schizophrenia, bipolar disorder and a healthy control group. The BADE task required rating four interpretations of delusion-neutral scenarios three times (in sequence) as increasingly disambiguating information was presented. A principal component analysis (PCA) carried out on all measures determined that two independent cognitive processes appear to combine to determine all responses on the BADE task: Integration of Evidence and Conservatism, with only the former discriminating between the severely delusional schizophrenia group and all other groups. Thus, integration of evidence appears to be functioning sub-optimally in severely delusional schizophrenia patients, resulting in a bias against disconfirmatory evidence (BADE). The cognitive process theorized to be underlying this effect is hypersalience of evidence-hypothesis matches.

GeneE: Gene and protein query expansion with disambiguation

When referring to genes, authors often use synonyms instead of the official gene symbols. In order to accurately retrieve as many relevant documents as possible, we have developed GeneE, a web application that expands a gene query to include all known synonyms, and adds disambiguation information for ambiguous terms, before forwarding the query to either PubMed, Google or Jane. The query expansion algorithm is also available as a web service. http://biosemantics.org/geneE

Repeated text in unrelated passages: Repetition versus meaning selection effects

Despite previous findings, Klin, Ralano, and Weingartner (2007) found transfer benefits across unrelated passages. After processing an ambiguous phrase in Story A that was biased toward its sarcastic meaning, readers were more likely to interpret the identical phrase in Story B as sarcastic, even though it contained no disambiguating information. In the present experiments, we found both repetition effects (a benefit for the lexical items) and meaning selection effects (a benefit for the selected meaning of the phrase) with short delays between Stories A and B; with longer delays, only repetition effects were found. Whereas decreasing the elaboration of the phrase eliminated both effects, moving the disambiguating context from before to after the phrase eliminated meaning selection effects only. We conclude that meaning selection effects, which are based on conceptual overlap, are more sensitive to context changes and less robust than repetition effects, which are based on both perceptual and conceptual overlap.

Biomedical word sense disambiguation with ontologies and metadata: automation meets accuracy

BackgroundOntology term labels can be ambiguous and have multiple senses. While this is no problem for human annotators, it is a challenge to automated methods, which identify ontology terms in text. Classical approaches to word sense disambiguation use co-occurring words or terms. However, most treat ontologies as simple terminologies, without making use of the ontology structure or the semantic similarity between terms. Another useful source of information for disambiguation are metadata. Here, we systematically compare three approaches to word sense disambiguation, which use ontologies and metadata, respectively.ResultsThe 'Closest Sense' method assumes that the ontology defines multiple senses of the term. It computes the shortest path of co-occurring terms in the document to one of these senses. The 'Term Cooc' method defines a log-odds ratio for co-occurring terms including co-occurrences inferred from the ontology structure. The 'MetaData' approach trains a classifier on metadata. It does not require any ontology, but requires training data, which the other methods do not. To evaluate these approaches we defined a manually curated training corpus of 2600 documents for seven ambiguous terms from the Gene Ontology and MeSH. All approaches over all conditions achieve 80% success rate on average. The 'MetaData' approach performed best with 96%, when trained on high-quality data. Its performance deteriorates as quality of the training data decreases. The 'Term Cooc' approach performs better on Gene Ontology (92% success) than on MeSH (73% success) as MeSH is not a strict is-a/part-of, but rather a loose is-related-to hierarchy. The 'Closest Sense' approach achieves on average 80% success rate.ConclusionMetadata is valuable for disambiguation, but requires high quality training data. Closest Sense requires no training, but a large, consistently modelled ontology, which are two opposing conditions. Term Cooc achieves greater 90% success given a consistently modelled ontology. Overall, the results show that well structured ontologies can play a very important role to improve disambiguation.AvailabilityThe three benchmark datasets created for the purpose of disambiguation are available in Additional file 1.

Context-dependent interpretation of words: Evidence for interactive neural processes

The meaning of a word usually depends on the context in which it occurs. This study investigated the neural mechanisms involved in computing word meanings that change as a function of syntactic context. Current semantic processing theories suggest that word meanings are retrieved from diverse cortical regions storing sensory–motor and other types of semantic information and are further integrated with context in left inferior frontal gyrus (LIFG). Our fMRI data indicate that brain activity in an area sensitive to motion and action semantics – the posterior middle temporal gyrus (PMTG) – is modulated by a word’s syntactic context. Ambiguous words such as bowl were presented in minimal disambiguating contexts indicating object (the bowl) or action (to bowl) meanings and were compared to low-ambiguity controls. Ambiguous words elicited more activity than low-ambiguity controls in LIFG and various meaning-related areas such as PMTG. Critically, ambiguous words also elicited more activity in to-contexts than the-contexts in PMTG and LIFG, suggesting that contextual integration strengthened the action meaning in both areas. The pattern of results suggests that the activation of lexical information in PMTG was sensitive to contextual disambiguating information and that processing context-dependent meanings may involve interactions between frontal and posterior areas.

Modeling Task-Oriented Dialogue

A common tool for improving the performance quality of natural language processing systems is the use of contextual information for disambiguation. Here I describe the use of a finite state machine (FSM) to disambiguate speech acts in a machine translation system. The FSM has two layers that model, respectively, the global and local structures found in naturally-occurring conversations. The FSM has been modeled on a corpus of task-oriented dialogues in a travel planning situation. In the dialogues, one of the interactants is a travel agent or hotel clerk, and the other a client requesting information or services. A discourse processor based on the FSM was implemented in order to process contextual information in a machine translation system. Evaluation results show that the discourse processor is able to disambiguate and improve the quality of the dialogue translation. Other applications include human-computer interaction and computer-assisted language learning.