segmentation: at first segment the query into words.
query-analysis: find what kind of query it asks via RULES.
lexical-answer-type: find the key word of the query, i.e. what does it want.
query-reformulation:
data-process
data-parse: extract the text or table from the raw-data.
information-extraction: extract the effictive information from text.
answer-retrivel:
data-retrivel: given a query, return a list of related words/sentence.
text inference: no deep learning yet.
candidata-answer-rank: rank the answers.