Papers by Akanksha Bindal
In the current scenario, with the transpiring big data explosion, data sets are often too large t... more In the current scenario, with the transpiring big data explosion, data sets are often too large to fit completely inside the computers ´ internal memory. In efficient processes, speed is not an option, it is a must. Hence every alternative is explored to further enhance performance, by expanding in-place memory storage that enables more data to be resident in the memory, eliminating operation latency, and even deploying an in-memory database (IMDB) system where all the data can be kept in memory. However, the technique of in-memory data handling is still at an infant stage and not viable in the current scenario. To tackle this problem a hierarchical hashing scheme is discussed where only one component of a big data structure resides in the memory. In this paper two data structures are explored: 1) Map which is implemented as self-balancing binary search trees or more commonly Red Black Trees and 2) Unordered Map which is based on hashing with chaining technique. Serialization and de...
ArXiv, 2021
When interacting with smart devices such as mobilephones or wearables, the user typically invokes... more When interacting with smart devices such as mobilephones or wearables, the user typically invokes a virtual assistant (VA) by saying a keyword or by pressing a button on the device. However, in many cases, the VA can accidentally be invoked by the keyword-like speech or accidental button press, which may have implications on user experience and privacy. To this end, we propose an acoustic false-trigger-mitigation (FTM) approach for on-device device-directed speech detection that simultaneously handles the voice-trigger and touch-based invocation. To facilitate the model deployment on-device, we introduce a new streaming decision layer, derived using the notion of temporal convolutional networks (TCN) [1], known for their computational efficiency. To the best of our knowledge, this is the first approach that can detect device-directed speech from more than one invocation type in a streaming fashion. We compare this approach with streaming alternatives based on vanilla Average layer, ...
ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Generating natural, diverse, and meaningful questions from images is an essential task for multim... more Generating natural, diverse, and meaningful questions from images is an essential task for multimodal assistants as it confirms whether they have understood the object and scene in the images properly. The research in visual question answering (VQA) and visual question generation (VQG) is a great step. However, this research does not capture questions that a visually-abled person would ask multimodal assistants. Recently published datasets such as KB-VQA, FVQA, and OK-VQA try to collect questions that look for external knowledge which makes them appropriate for multimodal assistants. However, they still contain many obvious and common-sense questions that humans would not usually ask a digital assistant. In this paper, we provide a new benchmark dataset that contains questions generated by human annotators keeping in mind what they would ask multimodal digital assistants. Large scale annotations for several hundred thousand images are expensive and time-consuming, so we also present an effective way of automatically generating questions from unseen images. In this paper, we present an approach for generating diverse and meaningful questions that consider image content and metadata of image (e.g., location, associated keyword). We evaluate our approach using standard evaluation metrics such as BLEU, METEOR, ROUGE, and CIDEr to show the relevance of generated questions with human-provided questions. We also measure the diversity of generated questions using generative strength and inventiveness metrics. We report new state-of-the-art results on the public and our datasets.
International Journal of Computer Applications, 2015
In the current scenario, with the transpiring big data explosion, data sets are often too large t... more In the current scenario, with the transpiring big data explosion, data sets are often too large to fit completely inside the computers´ internal memory. In efficient processes, speed is not an option, it is a must. Hence every alternative is explored to further enhance performance, by expanding in-place memory storage that enables more data to be resident in the memory, eliminating operation latency, and even deploying an in-memory database (IMDB) system where all the data can be kept in memory. However, the technique of in-memory data handling is still at an infant stage and not viable in the current scenario. To tackle this problem a hierarchical hashing scheme is discussed where only one component of a big data structure resides in the memory. In this paper two data structures are explored: 1) Map which is implemented as self-balancing binary search trees or more commonly Red Black Trees and 2) Unordered Map which is based on hashing with chaining technique. Serialization and deserialization operations are also performed to free the internal memory and preserve the data structure object for later use. Operations such as read, write are performed, along with documentation of the results and illustrations of visual representations of the two algorithmic
This paper brings out the semantic search performance of traditional search engines and semantic-... more This paper brings out the semantic search performance of traditional search engines and semantic-based search engines.
Initially, four traditional search engines (Yahoo, Yandex,
Dogpile and Google) and two semantic search engines (Bing,
Kngine) are selected to compare their search performance
on the basis of precision ratio and how they handle natural
language queries. Twelve queries, from various topics were
run on each search engine, the first thirty documents on
each retrieval output was classified as being relevant or nonrelevant.
Afterwards, precision ratios were calculated for the
first 30 document retrieved to evaluate performance of these
search engines.Five natural language queries are then run on
each of the aforementioned search engines to measure the
relevancy of each retrieved document. These documents were
classified as relevant or not relevant. The semantic search
engines tend to handle these natural language queries with
up to 70-90 percent precision while the traditional search
engines fail to handle these natural language queries. Also,
the authors inferred that Google is a traditional engine which
is developing its semantic base at a rapid rate to be counted
amongst the semantic engines in the near future.
Uploads
Papers by Akanksha Bindal
Initially, four traditional search engines (Yahoo, Yandex,
Dogpile and Google) and two semantic search engines (Bing,
Kngine) are selected to compare their search performance
on the basis of precision ratio and how they handle natural
language queries. Twelve queries, from various topics were
run on each search engine, the first thirty documents on
each retrieval output was classified as being relevant or nonrelevant.
Afterwards, precision ratios were calculated for the
first 30 document retrieved to evaluate performance of these
search engines.Five natural language queries are then run on
each of the aforementioned search engines to measure the
relevancy of each retrieved document. These documents were
classified as relevant or not relevant. The semantic search
engines tend to handle these natural language queries with
up to 70-90 percent precision while the traditional search
engines fail to handle these natural language queries. Also,
the authors inferred that Google is a traditional engine which
is developing its semantic base at a rapid rate to be counted
amongst the semantic engines in the near future.
Initially, four traditional search engines (Yahoo, Yandex,
Dogpile and Google) and two semantic search engines (Bing,
Kngine) are selected to compare their search performance
on the basis of precision ratio and how they handle natural
language queries. Twelve queries, from various topics were
run on each search engine, the first thirty documents on
each retrieval output was classified as being relevant or nonrelevant.
Afterwards, precision ratios were calculated for the
first 30 document retrieved to evaluate performance of these
search engines.Five natural language queries are then run on
each of the aforementioned search engines to measure the
relevancy of each retrieved document. These documents were
classified as relevant or not relevant. The semantic search
engines tend to handle these natural language queries with
up to 70-90 percent precision while the traditional search
engines fail to handle these natural language queries. Also,
the authors inferred that Google is a traditional engine which
is developing its semantic base at a rapid rate to be counted
amongst the semantic engines in the near future.