Abstract

Language identification (LID) is vital frontend for spoken dialogue systems operating in diverse linguistic settings to reduce recognition and understanding errors. Existing LID systems which use low-level signal information for classification do not scale well due to exponential growth of parameters as the classes increase. They also suffer performance degradation due to the inherent variabilities of speech signal. In the proposed approach, we model the language-specific phonotactic information in speech using recurrent neural network for developing an LID system. The input speech signal is tokenized to phone sequences by using a common language-independent phone recognizer with varying phonetic coverage. We establish a causal relationship between phonetic coverage and LID performance. The phonotactics in the observed phone sequences are modeled using statistical and recurrent neural network language models to predict language-specific symbol from a universal phonetic inventory. Proposed approach is robust, computationally light weight and highly scalable. Experiments show that the convex combination of statistical and recurrent neural network language model (RNNLM) based phonotactic models significantly outperform a strong baseline system of Deep Neural Network (DNN) which is shown to surpass the performance of i-vector based approach for LID. The proposed approach outperforms the baseline models in terms of mean F1 score over 176 languages. Further we provide significant information-theoretic evidence to analyze the mechanism of the proposed approach.

Abstract

Thispaperpresentsourworktoapplynonlinearneuralnetworkforparsingfiver esourcep oorI ndianL anguagesbe-longingtotwomajorlanguagefamilies-Indo-AryanandDravidian.BengaliandMarathiareIndo-AryanlanguageswhereasKannada,TeluguandMalayalambelongtotheDravidianfamily.WhilelittleworkhasbeendonepreviouslyonBengaliandTelugulineartransition-basedparsing,wepresentoneofthefirstparsersforMarathi,KannadaandMalayalam.AlltheIndianlanguagesarefreewordorderandrangefrombeingmoderatetoveryrichinmorphology.Thereforeinthisworkweproposetheusageoflinguisticallymo-tivatedmorphologicalfeatures(suffixandpostposition)inthenonlinearframework,tocapturetheintricaciesofboththelan-guagefamilies.Wealsocapturechunkandgender,number,personinformationelegantlyinthismodel.Weputforwardwaystorepresentthesefeaturescosteffe

Abstract

Recent improvement in usage of digital media has led people to share their opinions about specific entity through audio. In this paper, an approach to detect the sentiment of an online spoken reviews based on relative prosody features is presented. Most of the existing systems for audio based sentiment analysis use conventional audio features, but they are not problem specific features to extract the sentiment. In this work, relative prosody features are extracted from normal and stressed regions of audio signal to detect the sentiment. Stressed regions are identified using the strength of excitation. Support Vector Machine (SVM) and Gaussian Mixture Model (GMM) classifiers are used to build the sentiment models. MOUD database is used for the proposed study. Experimental results show that, the rate of detecting the sentiment is improved with relative prosody features compared with the prosody and Mel Frequency Cepstral Coefficients (MFCC) because the relative prosody features has more sentiment specific discrimination compared to prosody features.

Abstract

Hindi and Urdu are spoken primarily in northern India and Pakistan; together, they constitute the third largest language spoken in the world. 1 They are two standard- ized registers of what has been called the Hindustani language, which belongs to the Indo-Aryan language family. Masica [1993] states that, while they are different languages officially, they are not even different dialects or subdialects in a linguistic sense. Rather, they are different literary styles based on the same linguistically defined

Abstract

We present a simple, fast and unsupervised approach for exploiting morphological regularities present in high dimensional vector spaces. We propose a novel method for generating embeddings of words from their morphological variants using morphological transformation operators. We evaluate this approach on MSR word analogy test set (Mikolov et al.,2013d) with an accuracy of 85% which is 12% higher than the previous best known system.

Abstract

A mismatch in training and operating environments causes a performance degradation in speech recognition systems (ASR). One major reason for this mismatch is due to the presence of expressive (emotive) speech in operational environments. Emotions in speech majorly inflict the changes in the prosody parameters of pitch, duration and energy. This work is aimed at improving the performance of speech recognition systems in the presence of emotive speech. This work focuses on improving the speech recognition performance without disturbing the existing ASR system. The prosody modification of pitch, duration and energy is achieved by tuning the modification factors values for the relative differences between the neutral and emotional data sets. The neutral version of emotive speech is generated using uniform and non-uniform prosody modification methods for speech recognition. During the study, IITKGP-SESC corpus is used for building the ASR system. The speech recognition system for the emotions (anger, happy and compassion) is evaluated. An improvement in the performance of ASR is observed when the prosody modified emotive utterance is used for speech recognition in place of original emotive utterance. An average improvement around 5% in accuracy is observed due to the use of non-uniform prosody modification methods.

Abstract

Code-Mixing (CM) is a natural phenomenon observed in many multilingual societies and is becoming the preferred medium of expression and communication in online and social media fora. In spite of this, current Question Answering (QA) systems do not support CM and are only designed to work with a single interaction language. This assumption makes it inconvenient for multi-lingual users to interact naturally with the QA system especially in scenarios where they do not know the right word in the target language. In this paper, we present WebShodh - an end-end web-based Factoid QA system for CM languages. We demonstrate our system with two CM language pairs: Hinglish (Matrix language: Hindi, Embedded language: English) and Tenglish (Matrix language: Telugu, Embedded language: English). Lack of language resources such as annotated corpora, POS taggers or parsers for CM languages poses a huge challenge for automated processing and analysis. In view of this resource scarcity, we only assume the existence of bi-lingual dictionaries from the matrix languages to English and use it for lexically translating the question into English. Later, we use this loosely translated question for our downstream analysis such as Answer Type(AType) prediction, answer retrieval and ranking. Evaluation of our system reveals that we achieve an MRR of 0.37 and 0.32 for Hinglish and Tenglish respectively. We hosted this system online and plan to leverage it for collecting more CM questions and answers data for further improvement.

Abstract

Word embeddings learned from text corpus can be improved by injecting knowledge from external resources, while at the same time also specializing them for similarity or relatedness. These knowledge resources (like WordNet, Paraphrase Database) may not exist for all languages. In this work we introduce a method to inject word embeddings of a language with knowledge resource of another language by leveraging bilingual embeddings. First we improve word embeddings of German, Italian, French and Spanish using resources of English and test them on variety of word similarity tasks. Then we demonstrate the utility of our method by creating improved embeddings for Urdu and Telugu languages using Hindi WordNet, beating the previously established baseline for Urdu.

Abstract

This paper presents DILTON, a system which solves simple arithmetic word problems. DILTON first predicts the operation that is to be performed (’-’,’+’,’*’,’/’) through a deep neural network based model and then uses it to generate the answer. DILTON divides the question into two parts - world state and query as shown in Figure 1. The world state and the query are processed separately in two different networks and finally the networks are merged to predict the final operation. DILTON learns to predict operations with 8.81 % in a corpus of primary school questions. With simple similarity between the contexts of quantities appearing in the problem and the question text, we are able to identify 92.25 % of relevant quantities and solve 81% of the questions. Our code and data is publicly available

Abstract

We present a language independent, unsupervised approach for transforming word embeddings from source language to target language using a transformation matrix. Our model handles the problem of data scarcity which is faced by many languages in the world and yields improved word embeddings for words in the target language by relying on transformed embeddings of words of the source language. We initially evaluate our approach via word similarity tasks on a similar language pair - Hindi as source and Urdu as the target language, while we also evaluate our method on French and German as target languages and English as source language. Our approach improves the current state of the art results - by 13% for French and 19% for German. For Urdu, we saw an increment of 16% over our initial baseline score. We further explore the prospects of our approach by applying it on multiple models of the same language and transferring words between the two models, thus solving the problem of missing words in a model. We evaluate this on word similarity and word analogy tasks.