ClickstreamSessionConverter Documentation¶
Overview¶
The ClickstreamSessionConverter
class is designed to convert clickstream session data into structured fine-tuning inputs, which incorporate query items, events, and associated metadata. This functionality streamlines the transformation of raw session data into a standardized representation suitable for model fine-tuning.
Functionality¶
The main method, convert
, processes session data, including search queries, results, and events, generating a structured format that can be used for model fine-tuning. The converter handles various input types, ensuring flexibility in adapting to different clickstream data formats.
Parameters¶
ClickstreamSessionConverter Class¶
item_type
: Class type for item metadata.query_item_type
: Class type for query items (default:TextQueryItem
).fine_tuning_type
: Class type for fine-tuning inputs (default:FineTuningInput
).event_type
: Class type for session events (default:DummySessionEventWithImportance
).
convert
Method¶
session_data
: ASessionWithEvents
object or a dictionary containing session attributes such as search query, search metadata, results, and events.
Returns¶
The convert
method returns a FineTuningInputWithItems
object that wraps a fine-tuning input with associated metadata items. This includes a query item and event importance scores.
Usage¶
The primary purpose of the ClickstreamSessionConverter
is to transform raw clickstream data into a structured input for model fine-tuning. It creates query items, filters events, and maps importance scores. Users can extend this base class to override conversion methods and customize the transformation logic according to specific project needs.
Example¶
An example session data dictionary might look like this:
session_data = {
"search_query": "example query",
"search_meta": {"key": "value"},
"search_results": [
{"object_id": 1, "meta": {"info": "sample"}},
{"object_id": 2, "meta": {"info": "sample2"}}
],
"events": [
{"object_id": 1, "event_importance": 0.8}
],
"created_at": "2021-01-01T00:00:00"
}
To use the converter:
converter = ClickstreamSessionConverter(ItemMeta)
fine_tuning_input = converter.convert(session_data)
This example demonstrates how to instantiate the ClickstreamSessionConverter
with the desired item metadata class and convert session data into a format ready for model fine-tuning.