Data: MultiDoc2Dial , by Song Feng, Siva Sankalp Patel, Hui Wan, Sachindra Joshi, CC BY 3.0 . The dataset's 488 help pages, each page's text taken from its HTML, with its headings marked by #s, its list items by -, and its blocks a blank line apart. Each page has its domain, and its title without the part number the dataset adds.
Data: MultiDoc2Dial , by Song Feng, Siva Sankalp Patel, Hui Wan, Sachindra Joshi, CC BY 3.0 . The opening question of each dialogue whose next turn answers it from the same page: 500 of the dataset's training dialogues, chosen at random, for validation, and 362 of its validation dialogues, for testing.
Data: MultiDoc2Dial , by Song Feng, Siva Sankalp Patel, Hui Wan, Sachindra Joshi, CC BY 3.0 . Each part of the page a question's answer was given from, as its first character in the page's text and the one after its last, moved onto the page's text from the dataset's by the characters other than space they share.
Data: MultiDoc2Dial , by Song Feng, Siva Sankalp Patel, Hui Wan, Sachindra Joshi, CC BY 3.0 . The pages cut into chunks six ways, each chunk as its page, its first character and the one after its last: the blocks joined in order until another would take a chunk past 100 or 200 words, windows of 100, 200 and 400 words, and windows of 200 words each starting 150 words after the one before.
Data: MultiDoc2Dial , by Song Feng, Siva Sankalp Patel, Hui Wan, Sachindra Joshi, CC BY 3.0 . The vector of each chunk of the blocks joined to 100 words, from nomic-embed-text-v1.5 at 8 bits, embedded after `search_document: `, one text a request, by llama.cpp's server, release b11734, and shortened to 256 numbers as the model's card shortens them, each number times 32,767, rounded.
Data: MultiDoc2Dial , by Song Feng, Siva Sankalp Patel, Hui Wan, Sachindra Joshi, CC BY 3.0 . The vector of each chunk of the blocks joined to 200 words, from nomic-embed-text-v1.5 at 8 bits, embedded after `search_document: `, one text a request, by llama.cpp's server, release b11734, and shortened to 256 numbers as the model's card shortens them, each number times 32,767, rounded.
Data: MultiDoc2Dial , by Song Feng, Siva Sankalp Patel, Hui Wan, Sachindra Joshi, CC BY 3.0 . The vector of each chunk of windows of 100 words, from nomic-embed-text-v1.5 at 8 bits, embedded after `search_document: `, one text a request, by llama.cpp's server, release b11734, and shortened to 256 numbers as the model's card shortens them, each number times 32,767, rounded.
Data: MultiDoc2Dial , by Song Feng, Siva Sankalp Patel, Hui Wan, Sachindra Joshi, CC BY 3.0 . The vector of each chunk of windows of 200 words, from nomic-embed-text-v1.5 at 8 bits, embedded after `search_document: `, one text a request, by llama.cpp's server, release b11734, and shortened to 256 numbers as the model's card shortens them, each number times 32,767, rounded.
Data: MultiDoc2Dial , by Song Feng, Siva Sankalp Patel, Hui Wan, Sachindra Joshi, CC BY 3.0 . The vector of each chunk of windows of 400 words, from nomic-embed-text-v1.5 at 8 bits, embedded after `search_document: `, one text a request, by llama.cpp's server, release b11734, and shortened to 256 numbers as the model's card shortens them, each number times 32,767, rounded.
Data: MultiDoc2Dial , by Song Feng, Siva Sankalp Patel, Hui Wan, Sachindra Joshi, CC BY 3.0 . The vector of each chunk of windows of 200 words sharing 50, from nomic-embed-text-v1.5 at 8 bits, embedded after `search_document: `, one text a request, by llama.cpp's server, release b11734, and shortened to 256 numbers as the model's card shortens them, each number times 32,767, rounded.
Data: MultiDoc2Dial , by Song Feng, Siva Sankalp Patel, Hui Wan, Sachindra Joshi, CC BY 3.0 . Each question's vector from nomic-embed-text-v1.5 at 8 bits, embedded after `search_query: `, one text a request, by llama.cpp's server, release b11734, and shortened to 256 numbers as the model's card shortens them, each number times 32,767, rounded.