opensearch semantic highlighter v1 EU AI Act Compliance Profile

opensearch-semantic-highlighter

Overview

The OpenSearch semantic highlighter is a trained classifier that takes a document and query as input and returns a binary score for each sentence in the document indicating its relevance to the query.

Usage

This model is intended to run inside an OpenSearch cluster. For production workloads you should deploy the traced version via the ML Commons plugin—see the OpenSearch documentation on semantic sentence highlighting models.

If you simply want to experiment outside a cluster you can run the source model locally. First install the dependencies (Python ≥ 3.8):

pip install torch transformers datasets nltk
python -m nltk.downloader punkt

Then run the example below:

import nltk
import torch
import numpy as np
from datasets import Dataset
from functools import partial
from torch.utils.data import DataLoader
from dataclasses import dataclass, field
from typing import Any, Dict, List, Union
from torch.nn.utils.rnn import pad_sequence
from transformers import AutoTokenizer, BertModel, BertPreTrainedModel
import torch.nn as nn

class BertTaggerForSentenceExtractionWithBackoff(BertPreTrainedModel):
    """Sentence-level BERT classifier with a confidence-backoff rule."""

    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels

        self.bert = BertModel(config)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)
        self.classifier = nn.Linear(config.hidden_size, self.num_labels)
        self.init_weights()

    def forward(
        self,
        input_ids=None,
        attention_mask=None,
        token_type_ids=None,
        sentence_ids=None,
    ):
        outputs = self.bert(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
        )

        se

Your obligations as a opensearch semantic highlighter v1 deployer

Track your progress

Sign up to track obligations, receive deadline reminders, and generate audit packages.

Tags

Languages

Library