Struct Lattice

pub struct Lattice { /* private fields */ }

Lattice which is constructed for performing the Viterbi search. Contain several parallel arrays. First level of parallel arrays is indexed by end word boundary. Word boundaries are always aligned to codepoint boundaries, not to byte boundaries.

During the successive analysis, we do not drop inner vectors, so the size of vectors never shrink. You must use the size parameter to check the current size and never access vectors after the end.

Implementations

impl Lattice

pub fn reset(&mut self, length: usize)

Prepare lattice for the next analysis of a sentence with the specified length (in codepoints)

pub fn connect_eos(&mut self, conn: &ConnectionMatrix<'_>) -> SudachiResult<()>

Find EOS node – finish the lattice construction

pub fn insert(&mut self, node: Node, conn: &ConnectionMatrix<'_>) -> i32

Insert a single node in the lattice, founding the path to the previous node Assumption: lattice for all previous boundaries is already constructed

pub fn connect_node( &self, r_node: &Node, conn: &ConnectionMatrix<'_>, ) -> (NodeIdx, i32)

Find the path with the minimal cost through the lattice to the attached node Assumption: lattice for all previous boundaries is already constructed

pub fn has_previous_node(&self, i: usize) -> bool

Checks if there exist at least one at the word end boundary

pub fn node(&self, id: NodeIdx) -> (&Node, i32)

Lookup a node for the index

pub fn fill_top_path(&self, result: &mut Vec<NodeIdx>)

Fill the path with the minimum cost (indices only). Attention: the path will be reversed (end to beginning) and will need to be traversed in the reverse order.

impl Lattice

pub fn dump<W: Write>( &self, input: &InputBuffer, grammar: &Grammar<'_>, lexicon: &LexiconSet<'_>, out: &mut W, ) -> SudachiResult<()>

Trait Implementations

impl Default for Lattice

fn default() -> Lattice

Returns the “default value” for a type.