Beyond storing the data, SocoDB reads and reasons about the text, and enables you to query the unstructured text as if they were structured data.
SocoDB makes it extremely easy to develop performant NLP-powered applications and trusted by the most innovative developers.
AI Database for Natural Language
More than just data storage, SocoDB reads the meaning of data.
Break it down
SocoDB first splits the text into a collection of spans: from sentences to entities.
Embed & structure
Each span is then annotated with neural embeddings and structured predictions.
Ensemble & link
Lastly, SocoDB creates an ensemble version of the data by discovering complex relations.
Query unstructured text as if they were structured
{
query: “What countries does X credit card
offer coupon for?”,
filters: {"creation_time": {"gte": "2019-03-05"}},
answer_type: "phrase"
}
{
results: [{
"answer": "North America, Japan, and South Korea",
"context": "CTrip holders have $200 coupon when
using the card at North America, Japan, and South Korea",
"src_doc": "2019-CTrip-Policy.docx"
}]
}
Retrieve information at any level of granularity with natural language
-
State-of-the-art open-domain question answering
-
A flexible DSL that supports both natural language and structured filters.
-
Pretrained deep-learning models that work out-of-box.
-
A real-time inference that scales to millions of documents.
Extract structured Information from the unstructured text
-
Limitless extraction via natural language-based extractors
-
Blazing fast extraction speed and human-level accuracy
-
The easiest way to convert from text data to structured tables
{
extract: {
"subject": "Bakery Square Apartment",
"extractors": {
"avg_rate": "What's monthly rate for {subj} 2B2B?",
"neighbor": "What's near {subj}?"
}}
}
{
results: {
"avg_rate": [{
"answer": "$2400/month",
"src_doc": "aparment.com.html",
}],
"neighbor": [{
"answer": "Google, UPMC, Anthropologie, LA Fitness",
"src_doc": "bakery-square.html",
}]}
}
{
volume: {
"query": “what do people like about restaurant X?”,
"size": 2
}
}
{
results: [{
"answer": "The service is really nice",
"hit_count": 4535,
"total_count": 14523,
"proportion": 0.312
},
{
"answer": "sashimi is fresh",
"hit_count": 2342,
"total_count": 14523,
"proportion": 0.161
}
}
Find accurate statics of unstructured opinions with a simple query
-
Discover the most salient voices from user-generated data
-
Fine-grained opinion discovery via free-formed questions
-
Accurately estimate the volume of each major opinions
-
Machine reading intelligence that understands natural language variations.
Use Cases
Key Features
AI search and question answering
The state-of-the-art neural information retrieval engine that enables real-time natural language question answering.
Semantic Aggregation
Real-time text analytics to analyze millions of words and discover valuable insights at scale.
Cloud-Native
We make it easy for you to run SocoDB on public cloud, private cloud, or somewhere in between.
Auto Structuralization
Extract structured information from millions of unstructured documents into any format that you specified.
Deep Learning-Native
SocoDB is built-in with state-of-the-art deep learning models for natural language processing.
Ease of Use
SocoDB provides easy-to-use SDKs in Python, RESTful API as well as other Java, C++, Go+ (coming soon)