key | value | description |
---|---|---|
method (required) | object | For this object’s parameters, see the following table. |
anchor | The Anchor parameter is optional for fields that use this method.If you specify an anchor and leave the Multimodal Engine unconfigured or configured with “region”: “automatic” then: - Sensible ignores the anchor if it’s present in the document. - Sensible returns nulls for the fields in this query group if the anchor isn’t present in the document.If you specify an anchor and configure the Multimodal Engine parameter’s region manually, then Sensible creates the prompt’s context relative to the anchor. |
key | value | description |
---|---|---|
id (required) | queryGroup | |
queries | array of objects | An array of query objects, where each extracts a single fact and outputs a single field. Each query contains the following parameters:id (required) - The ID for the extracted field. description (required) - A free-text question about information in the document. For example, “what’s the policy period?” or “what’s the client’s first and last name?”. For more information about how to write questions (or “prompts”), see Query Group extraction tips. |
chunkScoringText | string | Configures context’s content. For details about context and chunks, see the Notes section.A representative snippet of text from the part of the document where you expect to find the answer to your prompt. Use this parameter to narrow down the page location of the answer to your prompt. For example, if your prompt has multiple candidate answers, and the correct answer is located near unique or distinctive text that’s difficult to incorporate into your question, then specify the distinctive text in this parameter.If specified, Sensible uses this text to find top-scoring chunks. If unspecified, Sensible uses the prompt to score chunks.Sensible recommends that the snippet is specific to the target chunk, semantically similar to the chunk, and structurally similar to the chunk. For example, if the chunk contains a street address formatted with newlines, then provide a snippet with an example street address that contains newlines, like 123 Main Street\nLondon, England. If the chunk contains a street address in a free-text paragraph, then provide an unformatted street address in the snippet. |
multimodalEngine | object | Configure this parameter to: - Extract data from images embedded in a document, for example, photos, charts, or illustrations. - Troubleshoot extracting from complex text layouts, such as overlapping lines, lines between lines, and handwriting. For example, use this as an alternative to the Signature method, the Nearest Checkbox method, the OCR engine, and line preprocessors.This parameter sends an image of the document region containing the target data to a multimodal LLM (GPT-4 Vision Preview), so that you can ask questions about text and non-text images. This bypasses Sensible’s OCR and direct-text extraction processes for the region. Note that this option doesn’t support confidence signals.This parameter has the following parameters:region: The document region to send as an image to the multimodal LLM. Configurable with the following options : - To automatically select the context as the region, specify “region”: “automatic”. If you configure this option for a non-text image, then help Sensible locate the context by including queries in the group that target text near the image, or by specifying the nearby text in the Chunk Scoring Text parameter. - To manually specify a region relative to the field’s anchor, specify the region using the Region method’s parameters, for example:“region”: { “start”: “below”, “width”: 8, “height”: 1.2, “offsetX”: -2.5, “offsetY”: -0.25 } |
confidenceSignals | For information about this parameter, see Advanced prompt configuration. | |
contextDescription | For information about this parameter, see Advanced prompt configuration. | |
pageHinting | For information about this parameter, see Advanced prompt configuration. | |
chunkCount | default: 5 | For information about this parameter, see Advanced prompt configuration. |
chunkSize | default: 0.5 | For information about this parameter, see Advanced prompt configuration. |
chunkOverlapPercentage | default: 0.5 | For information about this parameter, see Advanced prompt configuration. |
pageRange | For information about this parameter, see Advanced prompt configuration. |
Example document | Download link |
---|
Example document | Download link |
---|
Example document | Download link |
---|
chunkScoringText
parameter. Sensible scores each chunk using the OpenAPI Embeddings API.4387-09-22-33
, Sensible matches the line Policy Number: 4387-09-22-33
in the document. Sensible implements fuzzy matching using Levenshtien distance.April 7
), Sensible chooses the top-scoring match. If candidates have similar scores, Sensible uses page location as a tie breaker and chooses the earliest match in the document.