-
Google introduced TyDi question and answer set that attempts to capture the idiosyncrasies and features of tongues.
-
TyDi QA is a set of questions and answers that contains 200k QA pairs from various languages.
-
Researchers conducted Google search to find a suitable question in various languages and asked people to highlight the answer from the same.
Google hopes to encourage the development of AI capable of understanding the ways in which languages express different meanings. To this end, company researchers detailed a data set — TyDi QA, a question-answering data set that covers 11 languages — inspired by typological variety, or the notion that different languages express meaning in structurally uncommon ways.
TyDi QA is one thing of a supplement to the Google launched the previous year, and it makes an attempt to seize the idiosyncrasies and lines of tongues like Jap and Arabic. The researchers indicate, as an example, that English adjustments phrases to suggest one object (“ebook”) as opposed to many (“books”), and that Arabic has a 3rd shape to suggest if there are two of one thing (“كتابان”, kitaban) past simply singular (“كتاب”, kitab) or plural (“كتب”, kutub).
As a result of we selected a set of languages that are typologically distant from each other for this corpus, we expect models performing well on this dataset to generalize across a large number of the languages in the world.
~ Jonathan Clark, Research Scientist at Google wrote in a blog post.
from languages representing a “diverse range” of linguistic phenomena and data challenges, many of which use non-Latin alphabets (such as Arabic, Bengali, Korean, Russian, Telugu, and Thai) and form words in complex ways (including Arabic, Finnish, Indonesian, Kiswahili, and Russian). The languages also range from those with an abundance of available data on the web (English and Arabic) to those with very little (Bengali and Kiswahili).
Learn more:
Unlike Google Assistance, The questions in TyDi QA data set were collected from people who wanted an answer but still did not know the answer to avoid original questions that contained the same words as the answer. To inspire questions, the researchers showed taxpayers a Wikipedia passage written in their native language. This made the task easier and doable for them. Then they were asked to ask a question, any question provided it was not answered by the passage and really wanted to know the answer. For instance, “Does a passage on the ice make you think of ice lollies in summer? Great! Ask who invented the ice lollipops.”
It is important to note that the questions were written directly in each language, not translated, so many questions were different from those seen in the first English corpus. (For example, সফেদা ফল খেতে কেমন?, Or “What does the sapote taste like?”)
For each of the questions, the researchers conducted a Google search to find the most appropriate Wikipedia article in the appropriate language and asked a person to search and highlight the answer in that article. In some languages, they discovered that words were represented very differently in questions and answers, so differently that they expect the design of a system to successfully select an answer from a Wikipedia article is a challenge.
To track the progress of the community, they have established a leaderboard where participants can assess the quality of their machine learning systems.
We hope that this dataset will push the research community to innovate in a way that creates more useful systems of questions and answers for users around the world,
~ Jonathan Clark, Research Scientist at Google wrote in a blog post.
What is TyDi QA?
is a benchmark for information-seeking question-answering in typologically diverse languages. Google presents TyDi QA, a question-answering dataset covering 11 typologically diverse languages with 200K question-answer pairs. The languages of TyDi QA are diverse with regard to their typology — the set of linguistic features that each language expresses — such that we expect models performing well on this set to generalize across a large number of the languages in the world. It presents a quantitative analysis of the data quality and example-level qualitative linguistic analyses of observed language phenomena that would not be found in English-only corpora. To provide a realistic information-seeking task and avoid priming effects, questions are written by people who want to know the answer but don’t know the answer yet, and the data is collected directly in each language without the use of translation.
Learn more: