your mark down looks like well and most LLMs can handle i think but ...
an embdder at least the old good ones are not trained on markdown tables and dont find any semantic relations
my tests show that json-format is the right way to have a logic and semantic connection between headers and data cells.
example: headers only first row (same approach can be used with if top/left cell empty: headers are in top row and first column.
[
{
"Name": "Lena Ortiz",
"City": "Barcelona",
"Role": "Designer",
"Score": 88
},
{
"Name": "Marcus Chen",
"City": "Singapore",
"Role": "Data Analyst",
"Score": 92
},
{
"Name": "Ava Patel",
"City": "London",
"Role": "Product Manager",
"Score": 85
},
{
"Name": "Jonas Müller",
"City": "Berlin",
"Role": "Engineer",
"Score": 90
}
]
your mark down looks like well and most LLMs can handle i think but ...
an embdder at least the old good ones are not trained on markdown tables and dont find any semantic relations
my tests show that
json-format is the right way to have a logic and semantic connection between headers and data cells.example: headers only first row (same approach can be used with if top/left cell empty: headers are in top row and first column.
[ { "Name": "Lena Ortiz", "City": "Barcelona", "Role": "Designer", "Score": 88 }, { "Name": "Marcus Chen", "City": "Singapore", "Role": "Data Analyst", "Score": 92 }, { "Name": "Ava Patel", "City": "London", "Role": "Product Manager", "Score": 85 }, { "Name": "Jonas Müller", "City": "Berlin", "Role": "Engineer", "Score": 90 } ]